Incentivizing Calculated Risk-Taking:Evidence from an Experiment with

Commercial Bank Loan Officers

SHAWN COLE, MARTIN KANZ, and LEORA KLAPPER∗

Journal of Finance, forthcoming

ABSTRACT

We conduct an experiment with commercial bank loan officers to test how perfor-

mance compensation affects risk-assessment and lending. High-powered incentives

lead to greater screening effort and more profitable lending decisions. This effect is,

however, muted by deferred compensation and limited liability, two standard features

of loan officer compensation contracts. We find that career concerns and personality

traits affect loan officer behavior, but show that the response to incentives does not

vary with traits such as risk-aversion, optimism or overconfidence. Finally, we present

evidence that incentive contracts distort the assessment of credit risk, even among

trained professionals with many years of experience.

JEL classification: D03, G21 J22, J33, L2Keywords: loan officer incentives, banking, emerging markets

∗Shawn Cole: Harvard Business School and NBER. Martin Kanz and Leora Klapper: The World Bank,Development Economics Research Group. We thank Campbell Harvey (the Editor), an Associate Editor, andan anonymous referee for their comments. We also thank Viral Acharya, Sumit Agarwal, Philippe Aghion,Andreas Fuster, Victoria Ivashina, Raj Iyer, Michael Kremer, Rohini Pande, Jose Liberti, Daniel Paravisini,Enrichetta Ravina, Andrei Shleifer, Antoinette Schoar, Erik Stafford, Jeremy Stein and numerous seminarand conference participants for helpful comments and suggestions. Samantha Bastian, Doug Randall, andWentao Xiong provided excellent research assistance. Financial support from the Kauffman Foundation,the International Growth Center, and Harvard Business School is gratefully acknowledged. The opinionsexpressed do not necessarily represent the views of the World Bank, its Executive Directors, or the countriesthey represent. Corresponding author: Martin Kanz, The World Bank, MC 3-368, 1818 H Street NW,Washington DC 20433, USA. Phone: +1 202-473-3943. Email: mkanz@worldbank.org.

I. Introduction

The effect of performance-based compensation on risk-taking is of fundamental importance

in finance.1 However, the precise mechanism through which financial incentives affect risk-

assessment and risk-taking remains poorly understood. Existing research has established

that bank lending is responsive to the external economic environment (see, for example,

Dell’Ariccia and Marquez [2006] or Keys et al. [2010])2 and, more recently, that agency

problems within a bank may affect credit allocation (Liberti and Mian [2009], and Hertzberg,

Liberti and Paravisini [2010]). By contrast, there is little evidence on individual responses to

performance-based compensation, a key instrument banks may use to influence the decisions

of loan officers tasked with making actual lending decisions.3

Linking compensation practices to lending decisions is difficult, for at least two impor-

tant reasons. First, incentive structures are endogenously determined by financial institu-

tions, yielding standard identification problems. Second, even setting identification chal-

lenges aside, the data typically available in observational studies, such as lending, interest

income, and write-offs, are often insufficient to distinguish between competing hypotheses.

To surmount these challenges, this paper uses a high-stakes field experiment4 with com-

mercial bank loan officers in India, which enables us to present direct evidence on the effect

of performance-based compensation on risk-assessment and lending decisions. In the exper-

1The impact of incentives on risk-taking has been cited as a key factor in many financial crises thatwere preceded by a lending boom. See Bebchuk, Cohen and Spamann [2010], Fahlenbrach and Stulz [2012],Acharya, Litov and Sepe [2013] for a discussion of incentives and risk-taking in the run-up to the recentglobal financial crisis. Devlin [1989] and Gourinchas, Valdes and Landerretche [2001] highlight the role ofemployee incentives and supply side factors in the Latin American debt crisis. For a general discussion ofincentives and risk-taking at banks, see also “Crazy compensation and the crisis”. Alan Blinder, The WallStreet Journal. May 28, 2009.

2For evidence on credit booms and screening incentives see also Dell’Ariccia, Igan and Laeven [2012].Theoretical approaches have modeled variation in screening standards as a result of herding, business cyclefactors (Kiyotaki and Moore [1997]), or limited screening capacity of banks (Berger and Udell [2004]).

3See Freixas and Rochet [2008] for a discussion of incentive problems specific to lending. For reviews ofincentive compensation in firms see Baker, Jensen and Murphy [1988] and Prendergast [1999].

4The design of our experiment combines elements of a field and lab experiment. We follow the classificationof proposed by Harrison and List [2004], who refer to this experimental design as a “framed field experiment”.

1

iment, loan officers were paid to review and assess actual loan applications, making 14,675

lending decisions under exogenously assigned incentives. We pinpoint the relationship be-

tween compensation and lending decisions by exogenously varying the incentive contracts

faced by loan officers and evaluate three classes of incentive schemes: (i) volume incentives

that reward origination, (ii) low-powered incentives that reward origination conditional on

performance and (iii) high-powered incentives that reward performance and penalize default.

While much of the literature on performance-based compensation in banking and finance

has focused on incentives for risk-taking provided to top management,5 this paper explores

the hypothesis, often advanced in the aftermath of the global financial crisis, that non-

equity incentives for loan originators, such as commissions, can play an important role in

determining the fate of a bank’s lending operation.6 Indeed, providing appropriate incentives

to employees at the lower tiers of a commercial bank’s corporate hierarchy is a difficult

problem: their very responsibility is to collect information that the bank cannot otherwise

observe, making monitoring difficult. They enjoy limited liability, and may have different

risk and time preferences than the bank’s shareholders.

The design of our experiment closely matches the loan approval process of low- docu-

mentation loans in an emerging credit market, and has several features that are particularly

well-suited for studying the question at hand. First, while still novel in finance, the use

of randomized experiments has grown rapidly in other areas of economics, in large part

because they allow clear tests of causal relationships. Our unique experimental approach,

which brings professional loan officers with many years of experience in credit assessment

into a controlled laboratory environment, allows us to track aspects of loan officer behav-

ior that would normally be unobservable to a bank or econometrician and allows for the

5See Jensen and Murphy [1990] and Murphy [1999] for an overview of this literature.6Acharya, Litov and Sepe [2013] note that the literature on executive compensation disagrees about the

effect of performance pay on risk-taking and argues that these conflicting results could be due to an omittedfactor, such as the impact of non-executive compensation on risk-taking.

2

causal identification of the impact of monetary incentives on loan officer behavior. By using

a population of experienced loan officers, our design also accounts for the important con-

cern that trained professionals may behave differently from non-professionals in controlled

experimental environments (see, for example, Palacios-Huerta and Volji [2008]).

Second, by design, our experiment focuses on the lending decision and allows us to

isolate the impact of performance pay on the quality of initial screening from other channels

that may affect lending, such as the collection of soft information or the degree of ex-post

monitoring.7 Finally, participants completed a set of standard psychological tests, similar to

those used in the literature on managerial characteristics and decision-making (Malmendier

and Tate [2005], Graham, Harvey and Puri [2013]). We use this information to shed light

on the mechanism through which incentives affect loan officer decisions, and to benchmark

the size of the effects. In particular, we examine whether monetary incentives affect lending

decisions directly or through their interaction with personality traits, such as overconfidence,

conscientiousness or risk-aversion.

We present three main results. Our first set of results documents the efficacy and limi-

tations of performance incentives in lending. We provide evidence that the structure of per-

formance incentives strongly affects screening effort, risk-assessment, and the profitability of

originated loans. Loan officers who are incentivized based on lending volume rather than

the quality of their loan portfolio originate more loans of lower average quality. By contrast,

high-powered incentives that reward loan performance and penalize bad lending decisions

cause loan officers to exert greater screening effort, reduce exposure to loans with higher

perceived ex-ante credit risk, and induce significantly more profitable lending decisions while

leading only to a small reduction in lending volume. Relative to a baseline treatment with

low-powered incentives, high-powered incentives increase the probability that a bad loan is

7The distinction between screening, information production and ex-post monitoring is also a feature ofthe real lending environment that is being replicated by our experiment, where these tasks are carried outby separate employees, each facing their own wage schedule.

3

detected and increase profits per originated loan by up to 3.5% of the median loan size; in

contrast, origination incentives lead to a substantial decline in the quality of originated loans

and reduce profits per loan by up to 5% of the median loan size. Although screening effort

is on average lower under pure volume incentives, it is worth noting that loan officers do not

indiscriminately approve all applications, which suggests the presence of career concerns or

reputational motivations.8

Building on these results, we explore a number of constraints, inherent to any incentive

contract in lending, that may limit the efficacy of pay for performance. Consistent with the

predictions of a simple model of loan officer decision-making, we find that deferred compen-

sation attenuates the effectiveness of high-powered incentives. When incentive payments are

awarded with a three-month delay, our measures of costly screening effort decline by between

5% and 14%, and we document a corresponding but less pronounced decline in the quality of

originated loans. Notably, we find that deferred compensation also moderates the negative

effect of incentive schemes that emphasize loan origination over the quality of originated

loans. Relaxing loan officers’ limited liability constraint (similar in spirit to giving a loan

officer equity in the loan) induces greater screening effort and leads to more conservative

lending decisions, but has only a moderate effect on the profitability of originated loans.

Second, we demonstrate that performance incentives have important effects on loan offi-

cers’ subjective perception of credit risk. We find that loan officers evaluating applications

under performance contracts that provide strong incentives for approval systematically in-

flate internal ratings they assign to the loans they process. While internal ratings are strongly

predictive of default under all incentive schemes, loan officers facing volume incentives inflate

risk ratings by as much as .3 standard deviations, irrespective of the underlying asset quality.

8To examine the role of non-monetary motivations in greater detail, Section IV. D explores the effect ofloan officer characteristics on screening behavior. We find evidence consistent with the presence of careerconcerns, and identify several personality traits that make a loan officer more likely to exert effort underany monetary incentive. At the same time, we document that personality traits do not amplify loan officers’response to monetary incentives.

4

Since incentives affect both risk ratings and approvals, the loan book approved under a per-

missive incentive scheme may therefore be of poorer quality but, based on internal ratings

alone, may in fact look less risky than a set of comparable loans approved under a more

conservative incentive contract.

Third, we provide evidence on the interaction between loan officer characteristics and

the response to performance incentives, using data from psychometric tests administered to

a subset of the participants in our experiment. We show that loan officer characteristics

have a strong effect on loan officers’ baseline level of screening effort. In particular, we find

evidence that career concerns are a key non-monetary determinant of loan officer behavior.

However, personality traits show only a weak interaction with monetary incentives. This

indicates that personality traits can be useful in identifying conscientious screeners, but are

unlikely to affect individual performance differentially under alternative incentive schemes.

This paper contributes to several literatures. A growing body of research highlights the

importance of incentives for the transmission and use of information in lending (Hertzberg,

Liberti and Paravisini [2010], Qian, Strahan and Yang [2011], Berg, Puri and Rocholl [2012]).

Most closely related to our study, Agarwal and Ben-David [2012] exploit a change in the

compensation structure of a U.S. bank and show that volume incentives lead to greater

risk-taking and a deterioration in loan performance.

Second, we contribute to the literature on incentive compensation and risk-taking. Ex-

isting research in this area has focused almost exclusively on risk-taking among CEOs and

senior management (see Bebchuk and Spamann [2010], Bolton, Mehran and Shapiro [2010]

and Fahlenbrach and Stulz [2012]). Mechanisms similar in their effect to equity compensa-

tion for senior executives have been proposed to align the incentives of employees at lower

levels of a bank’s corporate hierarchy with those of the bank.

Finally, our findings add to the literature on lending in informationally opaque credit

markets. We examine the role of loan officer effort and risk-assessment in an environment of

5

high idiosyncratic risk (see Petersen and Rajan [1994], Berger, Klapper and Udell [2001]).

This is related to, but distinct from, the special role played by loan officers in collecting soft

information, and monitoring borrowers following the disbursal of a loan.

While we feel that our setting offers important advantages –for example we are able

to study lending decisions amounting to the allocation of approximately US$ 88 million

in credit– there are also two limitations worth mentioning. First, this paper studies one

specific lending model, often used in practice, where the loan officer’s primary function is

to screen loans, rather than to prospect for new clients, cross-sell other products, or gather

soft information.9 This allows us to devise a clean test for the impact of incentives that can

rule out multitasking concerns, but naturally confines the scope of our analysis to the loan

officer’s traditional screening role (Freixas and Rochet [2008]). Second, while the information

environment and lending process in our experiment match what’s done in practice, one might

be concerned that studying lending decisions in a lab may lead us to underestimate the role

of career concerns and other longer term motivations that may influence behavior in a real

lending environment. As we shall see, we do find evidence of career concerns and other

reputational motivations. These should, however, be interpreted as lower bound estimates.

The remainder of the paper proceeds as follows. In Section II. we discuss the basic

incentive problem in lending. Section III. describes the experimental setting and design.

Section IV. reviews the empirical strategy and presents our results, and Section V. concludes.

9The organizational form of the lending process is a distinct topic that is being explored in concurrentwork. See for example Paravisini and Schoar [2012].

6

II. Performance Incentives in Lending

The potential for excessive10 and socially inefficient risk-taking in response to poorly

designed incentive schemes has long been recognized. However, in many real-world settings,

first-best contracts may be difficult to implement, as they require easily quantifiable criteria

against which to measure and reward performance. The basic incentive problem in lending

arises from the fact that loan officers are tasked with allocating the bank’s capital based

on private information and risk-assessments that are not independently verifiable by the

bank (Stein [2002]). This generates significant scope for agency conflict within the lending

institution and creates a strong rationale for the use of performance pay to align the risk

and time preferences of the bank’s employees with those of the institution.

There are, however, several important constraints that generally preclude a bank from

offering a first-best contract that would make a loan officer a fully liable residual claimant

of the loans she originates. First, loan officer effort is typically unobservable. Second, loan

officers are necessarily protected by limited liability, as they take decisions on large amounts

of money, which typically far exceed the amount of any penalty a bank could enforce to deter

bad lending decisions. Third, the risk and time preferences of loan officers are likely to differ

significantly from those of the bank’s shareholders. This may make it difficult to generate

effort with deferred pay conditioned on loan outcomes, rather than with an immediate bonus.

Finally, in a lending environment characterized by high aggregate and idiosyncratic risk, it

is difficult to reliably identify idiosyncratic defaults, which further complicates the use of

realized outcomes for the measurement of loan officer screening effort and performance.

Where banks provide performance incentives, loan officer compensation typically consists

10The literature does not provide a universally accepted definition of what constitutes “excessive” risk-taking. In the context of our experiment, we define excessive risk-taking with reference to the hypotheticalfirst-best contract in which a loan officer would be made a fully liable residual claimant of the loans sheoriginates. Excessive risk-taking denotes the case in which a loan officer with rational beliefs takes higherrisk at a weakly lower rate of return than she would under the hypothetical first-best contract.

7

of a fixed base salary plus a performance component. This performance component may place

weight on lending volume, loan performance, or a combination of the two.11

The debate on bank compensation has revolved around two main features of such incen-

tive contracts: first, the incentive power of the contract, which is a function of the reward

for good and the penalty for bad decisions, and determines the perceived cost of originating

a loan that might become delinquent. Second, the often short time-horizon of compensation,

which may lead loan officers to prioritize short-term gains over long-term loan performance.

Theoretical work on performance incentives in lending has been relatively scarce. Heider

and Inderst [2012] develop a model of relationship lending and analyze the optimal contract

when loan officers, in addition to their traditional screening role, also act as “salespeople”,

tasked with prospecting for new loans and producing soft information through the cultivation

of lending relationships. They show that in this setting, the optimal contract is a function of

the bank’s competitive position, as well as the degree of private information the loan officer

can conceal from her employer.

In contrast to this line of research, our experiment is set in a lending environment where

loan solicitation and approvals are strictly distinct, such that loan officers approve loans with

little or no contact with the borrower. This enables us to rule out multitasking concerns in

the response to incentives. An additional advantage of this approach is that we can rank

incentive contracts offered in the experiment with reference to the hypothetical first-best in

which a loan officer would be made a fully liable residual claimant of the loans she originates.

The separation of information collection and loan approvals is common for a wide range

of financial products, and especially prevalent in emerging markets where the small loan

11The U.S. Department of Labor, for example, describes the structure of loan officer compensation con-tracts as follows: “The form of compensation for loan officers varies. [...] Some institutions pay only salaries,while others pay loan officers a salary plus a commission or bonus based on the number of loans originated.”(See http://www/bls.gov/oco/ocos018.htm, as also cited in Heider and Inderst [2012]). Examples of spe-cific compensation schemes that reward loan officers based on lending volume or loan performance are alsodiscussed in Berg, Puri and Rocholl [2012] and Paravisini and Schoar [2012].

8

sizes, relative to the high fixed cost of screening, often rule out the use of an expensive

relationship lending model that relies on repeated personal interaction with the client. This

places greater importance on incentives at the time of the initial screening decision, which

is the focus of our analysis.

In this paper, we study the impact of performance pay in lending in the context of an

experiment with commercial bank loan officers in India. The design of our experiment builds

on a simple model of loan officer decision making, outlined in the Internet Appendix,12 in

which loan officer behavior depends on both financial incentives and non-monetary reputa-

tional concerns. Specifically, we assume that in addition to monetary rewards, loan officers

care about the possibility that their actions may affect others’ inference about their type.

This is the standard approach suggested by a growing literature on how to model behavior

in experiments with real subjects (see Harrison and List [2004] and Levitt and List [2007]).

Our theoretical framework makes four basic predictions about the effect of performance

pay on loan officer behavior: first, origination incentives, as often employed by commercial

banks, lead to indiscriminate lending, low effort and high defaults. By contrast, high-powered

incentives that reward profitable lending and penalize default result in greater screening

effort, but more conservative lending. Second, deferred compensation reduces the power of

performance-based incentives. Third, relaxing a loan officer’s limited liability constraint, for

example through a contract with a “claw-back” provision, unambiguously increases effort.

Finally, effort under any contract is higher, and may be independent of monetary rewards,

12We follow, in particular, Levitt and List [2007], who propose a model in which an experimental subjectoptimizes a utility function that is additively separable in the monetary and non-monetary arguments: “Thechoice of action affects the agent through two channels. The first effect is on the individual’s wealth [. . . ], thesecond effect is the non-pecuniary moral cost or benefit associated with [the] action. [. . . ] More generally,we have in mind that decisions which an individual views as immoral, anti-social or at odds with her ownidentity (Akerlof and Kranton(2000, 2005)) may impose important costs on the decision maker)”. Thismodel is sufficiently general to encompass a range of non-monetary motivations including career concerns,the desire for social status and more general reputational motivations. See also Prendergast [1999] and Bloomand Van Reenen [2011] for evidence from the literature on personnel economics, and Bandiera, Barankay andRasul [2011] on non-monetary incentives in field experiments with firms using real employees as subjects.

9

if loan officers have reputational concerns.

III. Experimental Context and Design

A. Setting

We designed a ‘framed’ field experiment that closely matches the underwriting process

for unsecured small enterprise loans in India. In the experiment, loan officers recruited from

the active staff of several commercial banks evaluate credit applications in the context of a

controlled lab experiment under exogenously assigned incentives.

The files assessed in the experiment consist of real, previously processed loan applications.

Each file contains all information available to the bank at the time that the loan was first

evaluated. The distribution of loan files evaluated by loan officers in the experiment is

matched to the distribution of good and bad loans participants would expect to see in a real

lending environment. This distribution was elicited using a pilot survey of 30 loan officers

prior to the main experiment.

An especially attractive feature of this experimental design is that it allows us to draw on

a population of highly experienced loan officers and observe their behavior and decisions to a

level of detail that would be difficult to achieve outside a controlled laboratory environment.

In the analysis, we use this feature of the experiment to estimate the causal impact of alter-

native incentive schemes, as well as the interaction of monetary incentives with measurable

loan officer attitudes and personality traits. This allows us to provide causal evidence on the

channel through which incentives affect loan officer behavior.

While lending decisions in the experiment were hypothetical, in the sense that all loans

had been previously processed by a bank and their realized outcome had been observed,

loan officers received only information that was available to the bank at the original time of

10

application. Since we observe the performance of all evaluated loans, we were able to pay

participants performance incentives, based on their lending decision and the realized outcome

of the loan applications they approve. The experimental treatments vary the magnitude and

the time horizon of these conditional payments to change the terms of the incentive contract

faced by loan officers participating in the experiment.

One potential concern with our experimental design is that it might not fully account for

the role of soft information in loan giving. We note, however, that the aim of our experiment

is to isolate the impact of incentives on screening behavior and lending decisions. In order

to do this, we were careful to choose a loan product whose risk profile is determined by the

quality of the initial screening decision (rather than ex-post monitoring or soft information

obtained through relationship lending).

We focus on lending decisions for “mass market” loans to small businesses, as an example

of a loan product for which sales and origination channels are strictly distinct. Loans of this

type are sourced by sales agents in the field, who collect all necessary client information,

which is then forwarded to the bank’s loan officers for approval. Loan officers do not interact

with the client directly, cannot conduct interviews, and have no other way of collecting

soft information. By focusing on first-time borrowers, we remove the potential influence of

soft information generated over time. This allows for a clean test that isolates changes in

screening behavior from other channels through which incentives might affect lending. By

the same token, our results should be interpreted with care when applied to an environment

where loan officers and are incentivized on tasks beyond loan screening.

In order to ensure that monetary incentives in the experiment were perceived as salient,

we calibrated expected payouts to the approximate hourly wage of the median participant,

a public sector loan officer with ten years of experience in banking. The remainder of this

section describes the database of loans used in the experiment, the population of loan officers

and the experimental protocol.

11

B. Experimental Design

B.1 Loan Officers

Loan officers were recruited from the staff of several leading private and public sector

commercial banks in India. We report summary statistics for the population of participating

loan officers in Table I, columns [1] to [4]. The median loan officer in our sample is a public

sector bank employee who is 35 years old, and has 10 years of work experience. In Table I,

columns [5] to [8] we report comparable characteristics from a sample of all loan officers from

a major commercial bank in the region where our experiment takes place. The descriptive

statistics indicate that our sample is quite representative of this reference population in

terms of age, rank and experience. In addition to their participation in the experiment,

loan officers completed a series of tests of attitudes and personality traits, commonly used

in the literature on psychology and behavioral economics. Summary statistics of these tests

are reported in Panel B of Table I. Additional details on the measurement of loan officer

attitudes and personality traits are provided in the Internet Appendix.

[Place Table I about here]

B.2 Database of Loans

As a basis for the experiment, we requested a random sample of loan applications from

a large commercial lender in India and received 676 loan files. These loan files contain all

information available at the time the application was first processed, and are matched with

at least nine months of repayment history for each loan.13 The information contained in each

loan application can be grouped into the following categories, corresponding to the sections

of the Lender’s standard application format: (1) basic client information including a detailed

13More than 90% of all defaults occur during the first five months of a loan’s tenure, so that our defaultmeasure allows for a relatively precise measurement of loan quality.

12

description of the client’s business, (2) list of documents and verification, (3) balance sheet

and (4) income statement. In addition, participants in the experiment had access to three

types of background checks for each applicant: a site visit report on the applicant’s (5)

business and (6) residence and (7) a credit bureau report.14

Our sample consists of uncollateralized small business loans to self-employed individuals,

with a ticket size between Rs 150,000 (US$ 3,000) and Rs 500,000 (US$ 10,000).15 We

consider only term loans to new borrowers, many of whom are first-time applicants for a

formal loan.16 The median loan in our database has a tenure of 36 months, a ticket size of

Rs 283,214 (US$ 6,383) and a monthly installment of Rs 9,228. (US$ 208).

Based on the Lender’s proprietary data on loan repayment, we classify credit files into

performing and non-performing loans. Following the standard definition, we classify a loan

as delinquent if it has missed two or more monthly payments and remains 60+ days overdue.

To calculate the profitability of a loan, we subtract the disbursal amount from the discounted

stream of repayments.17 To achieve as representative a sample as possible, we also include a

subset of files from clients who applied, but were turned down by the Lender. Throughout

the analysis, we report results disaggregated by non-performing and declined loans and show

that our results are unaffected by the classification of loans declined ex-ante by the Lender.

Summary statistics for the sample of loan files are reported in the Internet Appendix,

Table D.II. The comparison between the sample of ex-post performing and non-performing

loans indicate that loan files indeed contain information that makes it possible to infer loan

quality, suggesting that there are returns to effort in this setting.

14We focus on loan applications from new customers. A credit bureau report was therefore only availablefor 66% of the loans in our sample.

15To rule out vintage effects and ensure consistency in the initial screening standards applied to loans usedin the experiment, we restrict our sample to loans originated in 2009 Q1 and 2009 Q2.

16Since none of the loans in our sample are collateralized, they are priced at an annual interest rate ofbetween 15 and 30 per cent. We control for the variation in interest rates by including loan fixed effects.

17We estimate the Lender’s net profit per loan as the net present value of the disbursal plus repaymentsincluding interest, discounted by 8%, the approximate rate on Indian commercial paper between January 1and December 31, 2009, and assuming a 10% recovery on defaulted loans.

13

B.3 Experimental Protocol

The experimental procedure and presentation of information were designed to closely

resemble the actual work environment of the representative loan officer.1819Incentive treat-

ments, as described in Section B.4 , were randomly and individually assigned at the loan

officer and session level. Loan officers were invited to an introductory session and then par-

ticipated in up to 15 sessions of the experiment, in which they evaluated a set of six randomly

assigned loans under a given incentive scheme. Within each session of the experiment, the

sequence of loan files was randomly assigned,20 but the ratio of performing, non-performing

and declined loans was held constant at four performing loans, one non-performing loan and

one loan declined by the Lender. We chose this ratio to match the distribution that loan

officers reported experiencing in their workplaces.

At the start of each session, loan officers were assigned to an incentive treatment, re-

ceived a one-on-one introduction to the incentive scheme in place and completed a short

questionnaire to verify comprehension. Loan officers then began the loan rating exercise in

which they were asked to assess a series of loan files, using a customized software interface.

For every loan file under review, the loan evaluation software reproduced each section of the

application on a separate tab on the loan officer’s screen: this included a description of the

applicant’s business, balance sheet, trade reference, site visit report, document verification

and credit bureau report when available. Each session of the experiment was scheduled to

last one hour, although participants could finish early or late if they so chose.

18Harrison, List and Towe [2007] point out that laboratory behavior may not match field behavior wheneliciting risk attitudes (“background risk”). In contrast to that study, we use within-subject variation,and the inclusion of loan officer fixed-effects may reduce the importance of heterogeneous perceptions ofbackground risk from different subjects.

19The literature on experiments in economics has pointed out that Hawthorne effects might obscure be-havior in experiments that occur under observation (see Levitt and List [2007] and Levitt and List [2011] fora discussion). Note that the only feature that changed in from session to session in our experiments was thecompensation scheme, so that any constant “experimenter demand” effects would not affect our estimates.

20This was done to ensure that estimates of loan performance would not be biased by factors such asvariation in the quality and extent of information contained in the application file.

14

While reviewing loan applications, participants were asked to assess the applicant’s credit

risk along 15 credit-scoring criteria adapted from the standard format of a leading Indian

bank. Internal ratings range from 0 to 100 (with a higher score indicating higher credit

quality) and were not binding for the loan officer’s lending decision. The risk ratings serve

three purposes. First, they add realism to the lab session, as completing a (non-binding)

risk rating is a routine part of evaluating applications. Second, they allow us to elicit a

measure of perceived credit risk that is not tied to loan officer compensation. Finally, internal

ratings serve to assist the loan officer in aggregating information about the application in a

systematic way. To ensure that internal ratings are an unbiased reflection of a loan officer’s

true risk-assessment, participants were reminded that internal risk ratings were not tied to

monetary incentives and never reviewed by the administrating staff.

Loan officers were asked to evaluate loans based on their best judgment, but were given

no information about the ratio of good and bad loans or the outcome of any particular loan

under evaluation.

B.4 Incentive Treatments

To test the impact of performance pay on loan officer behavior, we exogenously vary

three features of the incentive scheme faced by the loan officer: the incentive power of the

contract, the time horizon over which performance incentives are paid, and the degree of

limited liability enjoyed by a loan officer. We vary the power of the incentive contract by

assigning loan officers to contracts that specify three conditional payments: a payment wP

made when a loan is approved and performs, a payment wD, made when a loan is approved

and defaults and a payment w that is made when a loan is declined.

Because the outcome of a loan is only observed with some delay, performance incen-

tives, in practice, must be paid with a lag. In our setting, under the non-deferred payment

scheme, incentives were paid immediately following an experimental session. In the deferred

15

compensation scheme, incentive payments were delayed by three months.

Finally, we experimentally relax loan officers’ limited liability constraint, by providing

an initial endowment that the participant can lose if she approves non-performing loans.

This mimics proposed “clawback” schemes. Throughout the paper, we express experimental

incentive contracts as as the vector w = [wP,wD,w]. In addition to these three performance-

based conditional payments, loan officers received an unconditional show-up fee of Rs 100

(US$ 2.25) each time they participated in a session of the experiment.

In order to ensure that participants perceived these conditional payoffs as salient, we

calibrated the mean payout of experimental incentive schemes to approximately 1.5 times

the hourly wage of the median participant in our experiment, a public sector credit officer

with ten years of professional experience, an annual income of Rs 240,000 (US$ 4,800) and

an approximate hourly wage of Rs 125 (US$ 2.5).

Because understanding the impact of performance pay on costly screening effort is a main

objective of the experiment, half of our sessions included a “costly information” feature. In

this treatment condition, loan officers were given an initial information endowment of Rs

108. Under the costly information condition, loan officers were able to review only basic

client and loan information items for free21 and were charged Rs 3 per section for as many of

the remaining loan file sections as they chose to view. In these sessions, loan officers received

their remaining information endowment at the conclusion of the session, in addition to any

incentive payments. Table II summarizes the experimental incentive schemes.

[Place Table II about here]

We use the random assignment of incentive contacts to test the following predictions.

First, origination incentives will lead to greater risk-taking. Indeed, under this type of

incentive, purely rational and profit-maximizing loan officers should indiscriminately approve

21Two out of nine sections of the loan application could be viewed for free. This included the basiccustomer profile and the list of verified documentation provided.

16

all applications, and exert no effort to screen out bad applications.22 Second, high-powered

incentives will increase effort by increasing the rewards for a profitable lending decision and

increasing the penalty for originating a loan that ultimately becomes delinquent.23 Third,

high-powered incentives will induce more conservative lending behavior by increasing the

utility cost of making a bad lending decision. Fourth, if a loan officer’s discount rate is

greater than zero, the amount of effort induced by deferred compensation will be less than

the amount of effort induced by an immediate bonus.

Finally, if loan officers are intrinsically motivated, or responsive to reputational consider-

ations or career concerns, they may invest in screening even when such effort will not yield

additional remuneration.24

IV. Empirical Strategy and Results

A. Specification

Since treatment status was randomly assigned, our empirical strategy is straightforward

and we estimate regressions of the form:

yil =K−1∑k=1

βkTilk + θi + θl + ζ ′Ril + ξ′Xil + εil (1)

where yil is the outcome of interest for loan officer i and loan l, Til is a vector of treatment

dummies for the incentive schemes being compared to the baseline. In all regressions, we use

22It is of course possible that financial incentives interact with loan officers intrinsic motivation. If thisis the case, our experiments measure the combined effect of a “classical” increase in effort, along with anychanges in effort due to intrinsic motivation. This is the policy relevant parameter.

23Note that this implies that the effort exerted under these treatments can be ranked B > A > C.24Theoretical work has also suggested that monetary incentives may crowd out intrinsic motivation. How-

ever, a recurring theme in this literature is that some very restrictive conditions need to be fulfilled for“incentive crowding” to occur. Benabou and Tirole [2003] for example note that incentive crowding requiresthe employer (bank) to have an information advantage over the employee (loan officer). We believe that thisis unlikely to be the case in the setting we study.

17

the low-powered baseline incentive wB = [20, 0, 10] as the omitted category. We additionally

control for loan officer fixed effects, θi, loan file fixed effects θl, and individual controls Xil,

including loan officer age, seniority, rank, education, and dummies for whether the loan

officer has management or business experience. Finally, the experiments took approximately

one year to complete, and not all incentive schemes were eligible to be assigned in any

given session. Hence, our regressions include a set of fixed effects Ril to control for these

randomization strata. Standard errors are clustered at the loan officer-session level, the same

level at which the treatment is assigned.

Our dataset includes 14,369 lending decisions, representing 206 unique subjects, with

three key treatment conditions: (1) Low-powered incentives, which we use as the baseline

throughout the empirical analysis; (2) High-powered incentives, which reward loan officers

for approving loans that perform and penalizes the origination of loans that default; and (3)

Origination bonus, which rewards the loan officer for every originated loan.25

In addition to these incentive vectors, we vary conditions under which incentives are paid.

In 369 randomly selected sessions (2,214 loan evaluations), we defer incentive payments by 3

months, rather than paying immediately. In further 163 sessions (978 evaluations), we relax

the participant’s limited liability constraint by providing an initial information endowment

of Rs 200 (US$ 4.5), which can be lost if a loan officer makes a series of unprofitable lending

decisions. Finally, in 137 sessions (3,638 loans), we provide loan officers with an initial

information endowment of Rs 108 (US$ 2.25), which they may spend to sections of the loan

file. Table II summarizes the sample sizes by treatment condition. Table D.I in the Internet

Appendix reports a test of random assignment.

To test our hypotheses, we consider three primary groups of outcome variables: (i) mea-

sures of screening effort, (ii) measures of subjective risk-assessment, and (iii) lending decisions

25Regressions using all data we collected, which includes the performance bonus schemes which pay onlyif a loan performs, along with the appropriate treatment dummies, are reported in the Internet Appendix.

18

(actual risk-taking) and the resulting profitability of originated loans. We construct two mea-

sures of screening effort: the number of credit file sections reviewed by a credit officer; and

the amount of money spent on reviewing additional information under the costly-information

treatment. To measure risk-assessment and risk-taking, we record internal risk ratings as-

signed to each loan. Finally, to evaluate loan officer decisions and performance, we match the

loan officer’s lending decision to the actual profitability of the loan to the financial institution.

B. Descriptive Statistics – Loan Evaluations

Before turning to the main analysis, we report descriptive statistics of loan evaluations

during the exercise. We first verify that the experimental task is meaningful, in the sense

that it is indeed possible for loan officers to infer credit risk based on hard information

contained in an applicant’s loan file. Table D.II in the Internet Appendix presents mean

comparisons of loan application information for performing and non-performing loans. There

are a number of differences in hard information characteristics that help distinguish ex-post

performing from non-performing loans. In particular, borrowers who defaulted on their loans

had substantially lower revenue, younger businesses, higher ratios of monthly debt service

to income, compared to borrowers who remained current on their obligations. Overdues

on credit reports also predicted default. Higher-quality borrowers reported higher levels of

debt, consistent with the common observation of low-quality borrowers being excluded from

formal credit markets.

[Place Table III about here]

Table III reports summary statistics of loan evaluations by loan type and incentive.

We note the following. First, even for a group of highly experienced loan officers, making

profitable lending decisions in this lending environment was not a trivial task, as indicated by

the significant heterogeneity of performance across loan officers documented in Figure 1. On

19

average, loan officers approved 75% of all loans evaluated in the experiment and made correct

lending decisions in 64% of all cases. Lending volume responds dramatically to incentives.

Lending decisions were, however, profitable under all incentive schemes in the experiment

and would have earned the bank an average net present value of US$ 710 (11% of the

median loan size) per originated loan. Identifying performing loans was substantially easier

than identifying non-performing loans or loans that were rejected by the Lender ex-ante.

Changes in the incentive power of the contract were especially effective in improving loan

officer’s success in detecting non-performing loans, and these patterns are directly reflected

in the profitability of loans approved under alternative incentives.

[Place Figure 1 about here]

Table III, column [1] describes the number of sections a loan officer reviewed prior to

making a decision, while Table III, column [2] gives this number for only the subsample

which was charged to see additional sections from the loan file. Virtually all loan officers

study the basic information and borrower profile sections. However, some chose to reject or

accept a loan without viewing the entire application, particularly when the incentive scheme

did not reward higher-quality screening.26

In addition to observed lending decisions, we analyze loan officer risk assessment, as

measured by the rating each loan officer gave to each loan. Since ratings themselves were

not incentivized, one might wonder whether these ratings contain useful information. Fig-

ure 3 plots the distribution of loan officers’ risk ratings for performing and non-performing

loans and confirms that non-performing loans indeed received significantly lower ratings. A

Kolmogorov-Smirnov test rejects equality of the rating distributions at the 1% level.

Table D.IV in the Internet Appendix reports additional tests, in which we use internal

26When information was costly, loan officers were most likely to review sections of the loan file thatcontained basic financial information, such as income statements and balance sheet information, and muchless likely to pay for additional sections of the file such as site visit reports (results not reported in table).

20

ratings to predict loan approvals and performance. The results show that loan officer as-

sessments of credit risk are a meaningful and strongly significant predictor of actual lending

decisions, the probability of default and the profitability of loans. This is true for the overall

rating as well as its sub-components measuring personal and financial risk.

[Place Figure 2 about here]

Since loan officers complete multiple sessions, one might wonder whether loan officers

learn over the course of the study. An affirmative answer might be cause for concern, given

that our average loan officer has more than ten years of experience in lending. To verify

that learning over the course of the exercise poses no threat to the validity of our results,

Figure 2 plots the average fraction of correct decisions and average profit per originated loan

as a function of the number of completed sessions. These demonstrate no learning effect, a

result which is confirmed by a parametric test for learning during the experiment, reported

in Table D.III in the Internet Appendix.

[Place Figure 3 about here]

C. Results

C.1 Incentivizing Screening Effort

We first analyze the effect of incentives on screening effort. Intuitively, performance

incentives can affect the quality of lending decisions if they induce a loan officer to choose

higher screening effort, translating into a more thorough evaluation of available information.

The design of our experiment provides us with a straightforward measure of screening effort.

Specifically, we record how many of the ten sections of the credit file the loan officer chooses

to review before making a decision. In a separate set of sub-treatments meant to make the

effort trade-off even more stark, we charge loan officers Rs 3 for each section of the loan dossier

21

beyond what would be available on the application form.27 As human subject considerations

precluded an experimental design in which loan officers would pay to participate, we provide

each loan officer with an initial information endowment of Rs 108 (approximately US$ 2.25

per experimental session). Participants could choose not to pay for additional tabs, in which

case Rs 108 would be paid to them at the end of the session, in addition to whatever show-up

and incentive payments they earn. This information cost was not trivial: purchasing access

to all six tabs would cost close to the maximum payout of Rs 20 under the low-powered and

origination incentives. We use the amount spent to view loan sections as a second measure

of screening effort, capturing the notion of costly information. Because screening effort is

not observable to the bank, we do not tie bonus payments to measures of observed effort.

Table IV reports the effect of performance pay on screening effort, measured by the num-

ber of loan file sections reviewed when the only cost of effort was the loan officer’s time

(columns [1] and [2]), as well as when the loan officer was required to pay to view addi-

tional tabs (columns [3] and [4]). High-powered incentives significantly increase screening

effort. On average, loan officers facing high-powered incentives viewed .4 additional infor-

mation tabs when there was no charge to view additional sections (the mean number of tabs

viewed was 5.06 when information was free, and 3.99 when information was costly). When

information was costly, high-powered incentives had an even stronger effect, increasing the

average number of tabs viewed by .8-1.2. These effects are statistically significant across

all specifications. Interestingly, we do not observe effort to be significantly lower when loan

officers face origination bonuses, although the standard errors are not small enough to rule

out meaningful effects. These results confirm that loan officers respond strongly to monetary

incentives, and suggest that performance pay can incentivize effort in the review of borrower

information.

27Available for free were basic applicant details and list of provided documentation. Loan officers paid toview income statement, balance sheet, site visit reports, and trade and credit reference checks.

22

[Place Table IV about here]

C.2 Risk-Assessment and Risk-Taking

How do performance incentives affect the perception of credit risk and actual risk-taking?

We measure loan officers’ subjective risk assessment of credit risk using the non-binding

internal risk-ratings that participants were asked to complete while evaluating loans.

In Table V we use these internal ratings to explore the effect of incentives on the percep-

tion of credit risk. We find evidence that the structure of performance incentives distorts

the subjective assessment of credit risk. Loan officers facing incentives that reward loan

origination inflate internal ratings by as much as .16 standard deviations.

[Place Table V about here]

There are two interpretations that are consistent with this finding. Consider a model in

which loan officers screen to detect negative signals about a potential borrower. A reduction

in effort would result in fewer negative signals, and higher loan ratings. An alternative

possibility, which we cannot rule out, is that loan officers may fear harm to their reputation

if they approve a loan they have rated poorly, and therefore inflate ratings of loans they

are going to approve. Finally, our findings are also consistent with a behavioral view of

risk-assessment, which is outside the scope of our model. Loan officers may change their

perception of credit risk if they are not comfortable thinking that the loans they wish to

approve under prevailing incentives are indeed of poor quality. This “wishful thinking”

effect has been discussed extensively in connection with subprime lending in the United

States (see e.g. Barberis [2012]) and documented in lab experiments (Mayraz [2012]). While

our experiment does not allow us to disentangle the degree to which each of these forces

is at work, an important implication of our results is that irrespective of the underlying

mechanism, the same set of clients is judged as collectively less risky when the bank offers

23

an incentive scheme that places greater emphasis on lending volume.

We next turn to the effect of performance pay on risk-taking. Because the realized

outcome of a loan may be a poor proxy of the ex-ante riskiness at the time a loan is originated,

we take advantage of the fact that we had more than 100 loan officers rate each loan, and

construct two measure of ex-ante risk based on loan level ratings under the baseline incentive.

The first measure of risk is simply the loan’s average rating under the low-powered baseline

incentive. The second risk measure is the coefficient of variation, which measures the degree

of disagreement and uncertainty among loan officers about the riskiness of the loan.

If high-powered incentives encourage more discerning lending decisions, we would expect

loan officers to approve loans with higher average ratings and lower variance. Indeed, in

our data set, the coefficient of variation is strongly correlated with default. Table VI tests

this hypothesis. In the regressions in Table VI, we restrict the sample to loans which a loan

officer approved; thus the coefficients give the average risk rating of loans approved under

a particular incentive scheme. We find that high-powered incentives lead to slightly more

conservative lending intemrs of a loan’s average risk-rating, though this result is marginally

significant only for the measure of business and financial risk (Table VI, columns [5] and

[6]). However, we find strong evidence that high-powered incentives cause loan officers to

shy away from loans that are risky in the sense that there is greater ex-ante uncertainty

about the interpretation of information contained in the loan file, as reflected in greater

variance of a loan’s baseline risk rating. Loans approved under high-powered incentives are

characterized by a significantly lower coefficient of variation of their baseline rating (Table

VI, columns [7] to [12]).

[Place Table VI about here]

24

C.3 Lending Decisions and Loan-Level Profit

In Table VII, we turn to the impact of performance pay on lending decisions and loan

level profit. We find that loan officers facing compensation schemes that do not penalize

default are dramatically more likely to originate loans (Table VII, columns [1] and [2]).

Compared to the baseline condition, high-powered incentives lead to only slightly more

conservative lending decisions, with the share of loans approved dropping by between 3.6 and

.04%. This is a small effect relative to the mean acceptance rate of 71% under the baseline.

Incentive schemes that reward origination, on the other hand, result in a dramatic increase in

the probability of approval. Under the origination bonus treatment, loan approvals increase

by approximately 8 percentage points, statistically significant at the 1% level. Both results

are consistent with evidence from non-experimental studies of loan officer incentives (see, for

example, Agarwal and Ben-David [2012]).

[Place Table VII about here]

Of course, incentivizing more or less lending is relatively easy; the more interesting ques-

tion is whether incentives can make loan officers more discerning. Table VII, columns [3] and

[4] show that laxer incentives increase the fraction of good loan clients who are approved,

roughly in proportion to the overall effect on lending. We find a dramatically different pat-

tern for non-performing loans: loan officers facing the high-powered incentive scheme are 11

percentage points less likely to approve these bad loans, a result that is significant at the

5% level in column [5], despite the smaller sample size. In contrast, we find large increases

in the fraction of non-performing loans approved under an incentive scheme that does not

penalize poor screening decisions. The pattern is similar for the sample of loans rejected by

the bank, though the statistical significance of the high-powered incentive effect is lost.

In Table VII, columns [9] to [12], we study the effect of performance pay on the profitabil-

ity of bank lending. Our first measure is the net present value of repayments to the lender,

25

less the amount disbursed, restricting the sample to loans approved by our experimental sub-

jects.28 This measure is relevant for a capital-constrained lending institution that seeks to

maximize average profitability per loan made. columns [9] and [10] show that high-powered

incentives dramatically improve the profitability of lending, raising profit per loan by US$

149 to US$ 176 per loan, which correponds to approximately 3% of the median loan size.

The final two columns of Table VII consider profit per screened loan, setting the NPV of a

loan that is rejected by an experimental subject to zero. This measure makes most sense for

a lender whose lending opportunities may be limited and may face difficulties sourcing addi-

tional clients. Again, we find that high-powered incentives improve profitability by roughly

similar magnitudes, though the result is only statistically significant in the specification with

loan officer fixed-effects.

In our setting, the net interest margin is quite high (around 30%), so one might be con-

cerned that high-powered incentives lead loan officers to behave too conservatively, declining

profitable loans. In fact, we observe that high-powered incentives improve the quality of orig-

ination, and are therefore likely a profitable proposition from the bank’s perspective, even

when screening costs, reduced volume, and the cost of the incentive payments themselves

are taken into consideration.

C.4 Deferred Compensation

Efforts to regulate the compensation of loan originators have often focused on the alleged

“short-termism” present in many performance contracts in banking, and have therefore aimed

at extending the time-horizon of incentive payments. If loan officers have higher discount

rates than shareholders, however, deferred compensation will blunt the effect of incentives.29

28Because we do not observe the outcome of loans that were originally rejected by the lender, we do notinclude these loans in our profit calculations.

29One need not assume loan officers are impatient: credit-constraints or concern about separation fromemployers could also cause loan officers to discount future payments at high rates.

26

In this subsection, we test how the effects of incentive payments vary when the time

horizon of payouts is changed. It is worth noting that any compensation that varies with

loan repayment must be paid with some delay, as it takes time to observe whether loans

perform or not. The intent of our experimental treatments is to vary the extent of this

delay in performance-based compensation. We are primarily interested in understanding

whether deferred compensation weakens incentives for costly screening effort. We therefore

restrict attention to the subset of “costly information” treatments, in which loan officers

pay to access additional sections of the loan application. We operationalize the concept

of deferred compensation by comparing loan officer behavior under immediate performance

pay (for low-powered, high-powered and origination incentives) to behavior under a series of

treatments, in which incentive payments were awarded after a period of 90 days.30

Table VIII presents the results of the deferred compensation intervention. In Panel A, we

report the effect of deferred compensation on screening effort. Panel B reports on the effect

of deferred compensation on risk-taking, and treatment effects of deferred compensation

on loan-level profits are reported in Panel C. Note that in contrast to the previous tables,

the omitted category and relevant basis for comparison here is the low-powered treatment

with costly information. At the foot of the table, we report t-tests comparing the effect of

immediate versus deferred compensation. Consistent with basic theoretical predictions, the

results show that deferred compensation significantly weakens the impact of high-powered

incentives (Table VIII, columns [3] and [4]). This is most apparent in the effect of deferred

incentives on screening effort, as measured by loan sections purchased. In column [3], the

difference between immediate high-powered incentive payments and the exact same payments

deferred 90 days is large, [1.225 - (-.454)], and significant at the 1% level. While high-powered

30Note that our estimates do not differentiate between the pure effect of deferring incentive payments andthe lower real value of the payment at a future date. The setup of our treatments assumes that the relevantcomparison in a real world compensation contracts is between the nominal value of payment today versusthe same payment at a future date.

27

incentives induce loan officers to lend more conservatively (columns [5] and [6]), deferring

those same payments attenuates this effect. High-powered incentives lead loan officers to

shy away from loans that appear riskier ex-ante, irrespective of whether the high-powered

incentives are deferred (columns [7] and [8]). The point estimates of profitability are lower

for deferred weak (baseline) incentives, as well as the high-powered incentives, though the

difference is significant at the 10% level only for weak incentives. Finally, the results provide

some suggestive evidence that deferred compensation may mitigate some of the negative

implications of volume incentives. Although loan approvals are similarly high under deferred

and non-deferred origination incentives, loans approved under deferred origination incentives

are more profitable than loans approved under non-deferred volume incentives (Table VIII,

column [9]).

[Place Table VIII about here]

C.5 Relaxing Limited Liability

In the same way that banks benefiting from deposit insurance and other implicit guar-

antees may be tempted to take high-risk low-NPV gambles, loan officers seeking to maximize

their variable compensation may be tempted to take excessive risks due to the fact that they

are protected by limited liability. To test how the presence of limited liability, an inher-

ent characteristic of incentive contracts for loan originators, affects loan officer behavior,

we randomly assigned loan officers to a treatment that relaxed the officer’s limited liabity

constraint. In this treatment, participants received an endowment of Rs 200 (US$ 4.5) at

the beginning of each session, which was theirs to take home unless their incentive payments

for the session were negative. The worst outcome for a loan officer would be to approve two

bad loans and decline four good loans under high-powered incentives, in which case incentive

payments would be Rs -200 and the loan officer’s payout would be zero. The endowment

28

therefore completely relaxed the limited liability constraint for the session.

Table IX presents the results. We find evidence to suggest that relaxing limited liability

indeed increases loan officers’ screening effort (columns [3] and [4]), though the differences

are not statistically significant. Surprisingly, loan officers approve loans that appear to be

on average lower quality (column [5]) when limited liability is relaxed. When taking lending

decisions, loan officers are more conservative without limited liability, though the size of this

difference is modest (the difference in coefficients in column [7] is 2.9 percentage points) and

not statistically significant. Taken at face value, these results suggest that ensuring loan

officers have more skin in the game has only modest effects on effort and the profitability of

lending decisions. Note, however that in a loan officer’s real work environment “unlimited’

liability” may include career considerations and the possibility of losing one’s job. In our

experiment, we only partly capture these non-pecuniary factors, so that our results should

be interpreted as lower bound estimates.

[Place Table IX about here]

D. Do Loan Officer Characteristics Matter?

The analysis so far documents that the structure of performance pay has important

effects on loan officer behavior. However, individual ability and personality traits may play

an important role in determining how loan officers respond to incentives.

In this subsection, we use loan officer characteristics to explore the mechanisms by which

incentives affect behavior, and to compare their relative importance. We proceed in three

steps. First, we benchmark the effect of performance pay against the heterogeneity in per-

formance we observe absent such variation in monetary incentives. Second, we test whether

reputational motivations and career concerns can explain heterogeneity in effort, and doc-

ument the size of the effect. Finally, we examine whether financial incentives interact with

29

loan officer personality traits to determine screening behavior: are greater financial rewards,

for example, less effective in eliciting effort from risk-averse individuals?

We are able to answer these questions because our data collection efforts included a de-

tailed elicitation of loan officer characteristics and personality traits, including two widely

used personality tests: the ‘Big Five’ (BFI) personality test (John, Donahue and Kentle

[1991]) and the ‘LOT-R’ life orientation test (Scheier, Carver and Bridges [1994]). While

these tests are widely used, and a small literature has established that individual heterogene-

ity affects management decisions,31 there is little work that systematically links employee

personality traits to financial decisions. We are aware of only one study in the finance litera-

ture, Graham, Harvey and Puri [2013], which uses psychometric tests to link the personality

traits of senior executives to firms’ financial decisions.

We complement this work in several ways. We show that personality traits are an im-

portant determinant of employee behavior, and explain variation in effort, even in a setting

without explicit financial incentives. Second, an important finding in Graham, Harvey and

Puri [2013] is that growth firms employ less risk-averse executives, suggesting endogenous

matching between firm and employee. This, however, leaves open the question whether em-

poyees with different personality traits vary in their response to incentives. We are able

to address this question using exogenous variation in incentive contracts induced by our

experiment.

The most direct test for individual heterogeneity is a joint test of significance of the loan

officer fixed effects from equation (1). We reject the hypothesis that loan officer heterogene-

ity does not affect screening effort at the 1% level (F -Statistic 71.98, with N =204 degrees

of freedom). The magnitude of loan officer effects is economically significant, with officers

at the 75th percentile of the effort distribution viewing approximately 45% more tabs than

officers at the 25th percentile of the distribution. This significant heterogeneity in loan officer

31See Bertrand and Schoar [2003], Malmendier and Tate [2005] and Landier and Thesmar [2009].

30

effort suggests that the decision to exert screening effort depends on more than compensation

policy alone: even in settings without explicit monetary incentives, reputational concerns

and the prospect of promotion may motivate employees. Based on this observation, our next

test examines whether reputational concerns drive screening behavior, by examining whether

individuals whose characteristics indicate stronger reputational concerns behave differently.

To do this, we estimate regressions of the form:

yil =K−1∑k=1

βkT +L∑l=1

γlz +K−1∑k=1

L∑l=1

δkl(T ∗ z) + θi + θl + ζ ′Ril + ξ′Xil + εil (2)

where z is a personality trait, Xil is a control vector, which includes loan officer age, rank,

gender, education, business experience, dummy variables for branch manager experience and

employment at a private sector bank, Ril is a matrix of treatment conditions and all other

variables are as previously defined. We consider both the main effect of each personality

trait, as well as its interaction with the exogenously assigned monetary incentives. The

results are presented in Table X. dependent variable is always effort, which is measured by

the number of sections of the loan file that the loan officer reviews. Columns [1] to [4] include

all observations, while columns [5] to [8] are restricted to the sample of observations in which

loan offers faced an explicit monetary cost for viewing additional sections of the loan file.

[Place Table X about here]

In Table X, Panel A, we first consider the possibility that loan officer behavior is driven

by career concerns –a special type of reputational motivation, which would imply that effort

is a decreasing function of age. Consistent with the career concerns hypothesis we find that,

ceteris paribus, older loan officers exert less effort. Taking the point estimate from column

[5], a loan officer close to retirement (aged 60) will review .36 fewer loan file sections than a

30-year old officer. This represents a 10% reduction in effort. The presence of career concerns

31

can also explain why loan officers are motivated to exert effort and make appropriate lending

decisions even in the absence of explicit financial incentives.

The second entry of Table X, Panel A examines whether loan officers from private sector

banks behave differently than those employed by public sector banks. Private banks are likely

to be more meritocratic and offer faster promotion paths so that the returns to demonstrating

one’s type may be higher. Similarly, private banks may attract employees who are more

responsive to implicit career incentives. Both mechanisms would suggest stronger career

concerns, and imply that private bankers exert greater baseline effort when compared to

their public sector counterparts. We find that this is indeed the case: private sector loan

officers exert greater baseline screening effort under any monetary incentive scheme.

In TableX, Panel B, we data from psychometric tests and loan officer surveys to exam-

ine how loan officer behavior varies with fixed personality traits. We find that personality

matters: individuals who are risk-averse, altruistic, or state that they wish to live up to

personal and professional expectations exert significantly higher effort under any monetary

incentive. By contrast, loan officers who are overconfident32 screen significantly less. Per-

sonality also affects risk-ratings and the ability to correctly identify good loans. Optimistic

loan officers rate loans significantly higher. Risk-averse loan officers are significantly more

likely to approve non-performing loans while the opposite is true for impatient loan officers:

a loan officer in the top decile of the discount rate distribution is 16% more likely to originate

a non-performing loan than the average loan officer.33

Finally, to shed light on the mechanism through which performance-based compensation

affects loan officer behavior, we test whether monetary rewards affect effort and performance

directly, or through their interaction with fixed personality traits. Standard agency theory

would, for example, predict that it is more expensive to induce effort when agents are risk-

32We classify a loan officer as overconfident if she incorrectly ranks herself in the top decile of the perfor-mance distribution.

33The complete set of heterogeneous effects results are available in the Internet Appendix.

32

averse. The opposite may be true for traits such as optimism or overconfidence, which

might accentuate the response to high-powered incentives. The answer to this question

has important implications for financial firms: if the effects of incentives vary by employee

type, then firms must not only seek out employees with desirable personality traits, but also

consider which type of incentive contract is the best match for their employee population.

By and large, we find only weak evidence that the effects of incentives vary by personal

characteristics. In Table X, column [3] of Panel B.4, we find that more conscientious indi-

viduals alter their behavior less in response to changes in incentives, though this pattern is

not consistent across measures of effort (Table X, column [7] of Panel B.4). In total, five of

the twenty possible interactions in Table X, Panel B are statistically significant at the 5%

level. This is more than would be expected by chance, but does not provide overwhelming

evidence that incentives are mediated by personality type.

More confident, optimistic, or conscientious loan officers do not respond more strongly to

performance-based compensation than their peers. An interesting exception, consistent with

the reputational motivations documented above, is that private sector bankers respond to

incentives differently. In particular, we find an asymmetric response when officers are moved

from the baseline, low-incentive treatment. When given higher-powered incentives, private

sector officers do not increase effort, but when offered the origination bonus scheme, they

dramatically reduce effort.

In summary, our analysis suggests that career concerns are an important mechanism

which generates effort above and beyond what would be expected from immediate financial

incentives alone. A second lesson from this section is that personal characteristics are an

important determinant of loan officer behavior that may constrain the ability of performance

pay to affect screening effort. However, we do not find systematic evidence that incentives

work differentially for officers with different personal characteristics.

33

V. Conclusion

Understanding how performance compensation affects risk-taking is a question of first

order importance in finance. However, identifying the individual response to incentives is

difficult, as we rarely observe decisions under exogenously different incentive environments.

In this paper, we use an experiment with experienced loan officers to identify the effect of

performance-based compensation on risk-assessment and risk-taking. We find a strong and

economically significant effect of performance pay on risk-assessment and lending behavior.

Incentives that reward lending volume lead to high acceptance rates, low effort and high

default. By contrast, high-powered incentives are effective at generating effort, leading loan

officers to correctly identify and screen out bad loans, and raising the overall profitability

of lending. Giving loan officers “equity” in a loan that they can lose also leads to greater

effort, but does not appreciably improve the profitability of lending, as loan officers become

significantly more conservative and originate fewer loans.

At the same time, we document several factors that constrain the ability of conventional

incentive contracts to alter loan officer behavior. First, deferred compensation –a standard

feature of loan officer compensation contracts, due to the fact that loan outcomes are only

observed with some delay– severely attenuates the power of monetary incentives. In our

sample of professional loan officers, delaying bonus payments by just three months dramat-

ically reduces effort, and the profitability of lending. This important limitation may help

explain why we do not see front-line lending officers facing compensation which varies closely

with the performance of their loan portfolios. Interestingly, however, we find that deferred

incentives also limit the temptation to originate poor quality loans under permissive incen-

tive schemes. This suggests that extending the time horizon of loan officer compensation

can encourage more prudent lending decisions in settings where volume incentives are the

norm and where it may be difficult to implement pay-for-performance.

34

Second, using psychometric tests, we show that personality traits and demographic char-

acteristics have a strong effect on screening effort that is unrelated to monetary incentives.

Irrespective of monetary incentives, effort declines with age, which is consistent with the pres-

ence of career concerns. Our results suggest that performance pay affects behavior directly,

rather than by accentuating traits such as risk-aversion, conscientiousness or overconfidence.

A back-of-the-envelope calculation suggests that the effect of a loan officer’s personality type

is quantitatively important: under our baseline incentive scenario, where monetary incentives

are weak, a loan officer at the 75th percentile of the effort distribution reviewed 45% more

loan file sections than an officer at the 25th percentile of the distribution. This effect size is

large, in fact much larger than the increase in effort observed when loan officers transition

from baseline to high-powered incentives, which leads to a 23% increase in the number of

loan file sections reviewed. This variation helps explain why we observe screening effort even

when loan officers face no financial incentives. It also suggests that even in settings where

monetary incentives “work”, their efficacy may be bounded by fixed personality traits.

Finally, we provide evidence that monetary incentives distort the perception of credt risk:

permissive incentives lead loan officers to rate loans as significantly less risky than the same

loans evaluated under pay-for-performance.

These findings have important implications for the design of performance-based com-

pensation in lending. Lenders have increasingly relied on credit scoring models rather than

human judgment. But it is unclear whether credit scoring can outperform human judgment,

particularly in informationally opaque credit markets, such as the one we study. Nor is it

obvious what individual characteristics are associated with screening ability and to what

extent they help or hinder the use of performance incentives as a tool to manage credit-risk.

The results in this paper are a first step towards answering these important questions.

35

References

Acharya, Viral, Lubomir V. Litov and Simone M. Sepe. 2013. “Non-Executive Incentivesand Bank Risk-Taking.” Working Paper .

Agarwal, Sumit and Itzhak Ben-David. 2012. “Do Loan Officer Incentives Lead to LaxLending Standards?” Working Paper, Ohio State University, Fisher College of Business .

Akerlof, George A. and Rachel A. Kranton. 2000. “Economics and Identity.” QuarterlyJournal of Economics 115(3):715–753.

Ashraf, Nava, Dean Karlan and Wesley Yin. 2006. “Tying Odysseus to the Mast: Evi-dence from a Commitment Savings Product in the Philippines.” The Quarterly Journalof Economics 121(2):635–672.

Baker, George, Michael Jensen and Kevin Murphy. 1988. “Compensation and Incentives:Practice vs. Theory.” Journal of Finance 43(3):593–616.

Bandiera, Oriana, Iwan Barankay and Imran Rasul. 2011. “Field Experiments with Firms.”Journal of Economic Perspectives 25(3):63–82.

Barberis, Nicholas. 2012. Psychology and the Financial Crisis of 2007-2008. In FinancialInnovation and Crisis, ed. M. Haliassos. Cambridge, MA: MIT Press.

Bebchuk, Lucian, Alma Cohen and Holger Spamann. 2010. “The Wages of Failure.” YaleJournal on Regulation 27:257–282.

Bebchuk, Lucian and Holger Spamann. 2010. “Regulating Bankers’ Pay.” Georgetown LawJournal 98(2):247–287.

Benabou, Roland and Jean Tirole. 2003. “Intrinsic and Extrinsic Motivation.” The Reviewof Economic Studies 70(3):489–520.

Berg, Tobias, Manju Puri and Jorg Rocholl. 2012. “Loan Officer Incentives and the Limitsof Hard Information.” Working Paper, Duke University Fuqua School of Business .

Berger, Allen N. and Gregory F. Udell. 2004. “The Institutional Memory Hypothesis and theProcyclicality of Bank Lending Behavior.” Journal of Financial Intermediation 13(4):458–495.

Berger, Allen N., Leora F. Klapper and Gregory F. Udell. 2001. “The Ability of Banksto Lend to Informationally Opaque Small Businesses.” Journal of Banking & Finance25(12):2127–2167.

Bernheim, Douglas. 1994. “A Theory of Conformity.” Journal of Political Economy102(51):841–877.

36

Bertrand, Marianne and Antoinette Schoar. 2003. “Managing with style: the effect of man-agers on firm policies.” Quarterly Journal of Economics 118(4):1169–1208.

Bloom, Nicholas and John Van Reenen. 2011. Human Resource Management and Produc-tivity. In Handbook of Labor Economics, ed. Orley Ashenfelter and David Card. Vol. 4Bof Handbook of Labor Economics chapter 19, pp. 1697–1769.

Bolton, Patrick, Hamid Mehran and Joel Shapiro. 2010. “Executive Compensation andRisk-Taking.” Working Paper, Federal Reserve Bank of New York .

Dell’Ariccia, Giovanni, Deniz Igan and Luc Laeven. 2012. “Credit Booms and LendingStandards: Evidence from the Subprime Mortgage Market.” Journal of Money, Creditand Banking 44:367–384.

Dell’Ariccia, Giovanni and Robert Marquez. 2006. “Lending Booms and Lending Standards.”Journal of Finance 61(5):2511–2546.

Devlin, Robert. 1989. Debt and Crisis in Latin America: the Supply Side of the Story.Princeton: Princeton University Press.

Fahlenbrach, Rudiger and Rene Stulz. 2012. “Bank CEO Incentives and the Credit Crisis.”Journal of Financial Economics 99(1):11–26.

Freixas, X. and J.-C. Rochet. 2008. Microeconomics of Banking. Cambridge, MA: MITPress.

Gourinchas, Pierre-Olivier, Rodrigo Valdes and Oscar Landerretche. 2001. “Lending Booms:Latin America and the World.” Economı́a 1(2):47–99.

Graham, John, Campbell Harvey and Manju Puri. 2013. “Managerial Attitudes and Corpo-rate Actions.” Journal of Financial Economics. 109(1):103–121.

Harrison, Glenn W, John A List and Charles Towe. 2007. “Naturally Occurring Preferencesand Exogenous Laboratory Experiments: A Case Study of Risk Aversion.” Econometrica75(2):1468–1482.

Harrison, Glenn W. and John List. 2004. “Field Experiments.” Journal of Economic Liter-ature 42(4):1013–1059.

Heider, Florian and Roman Inderst. 2012. “Loan Prospecting.” Review of Financial Studies.25:2381–2415.

Hertzberg, Andrew, Jose Liberti and Daniel Paravisini. 2010. “Information and IncentivesInside the Firm: Evidence from Loan Officer Rotation.” The Journal of Finance 65(3):795–828.

37

Jensen, Michael and Kevin Murphy. 1990. “Performance Pay and Top-Management Incen-tives.” Journal of Political Economy 98(2):225–64.

John, Oliver P., E. Donahue and R. Kentle. 1991. “The Big Five Inventory–Versions 4a and54.” Berkeley, Institute of Personality and Social Research, Berkeley, CA: University ofCalifornia .

Keys, Benjamin J., Tanmoy K. Mukherjee, Amit Seru and Vikrant Vig. 2010. “Did Secu-ritization Lead to Lax Screening? Evidence from Subprime Loans.” Quarterly Journal ofEconomics 125(1):307–362.

Kiyotaki, Nobuhiro and John Moore. 1997. “Credit Cycles.” Journal of Political Economy105(2):211–48.

Landier, Augustin and David Thesmar. 2009. “Financial Contracting with Optimistic En-trepreneurs: Theory and Evidence.” Review of Financial Studies 22(1):117–150.

Levitt, Steven D. and John A List. 2007. “What do Laboratory Experiments MeasuringSocial Preferences Reveal about the Real World?” Journal of Economic Perspectives21(2):153–174.

Levitt, Steven and John List. 2011. “Was There Really a Hawthorne Effect at the HawthornePlant? An Analysis of the Original Illumination Experiments.” American Economic Jour-nal: Applied Economics 3(1):224–238.

Liberti, Jose and Atif Mian. 2009. “The Effect of Hierarchies on Information Use.” Reviewof Financial Studies 22(4):4057–4090.

Malmendier, Ulrike and G. Tate. 2005. “CEO Overconfidence and Corporate Investment.”Journal of Finance 60(6):2661–2700.

Mayraz, Guy. 2012. “Wishful Thinking.” Working Paper, Oxford University, Department ofEconomics .

Meier, Stephan and Charles Sprenger. 2010. “Present-Biased Preferences and Credit CardBorrowing.” American Economic Journal: Applied Economics 2(1):193–210.

Murphy, Kevin. 1999. Executive Compensation. Elsevier Science North Holland.

Palacios-Huerta, Ignacio and Oscar Volji. 2008. “Experientia Docet: Professionals PlayMinimax in Laboratory Experiments.” Econometrica 76(1):71–115.

Paravisini, Daniel and Antoinette Schoar. 2012. “The Information and Agency Effects ofScores: Randomized Evidence from Credit Committees.” Working Paper, London Schoolof Economics .

38

Petersen, Mitchell A. and Raghuram G. Rajan. 1994. “The Benefits of Lending Relationships:Evidence from Small Business Data.” The Journal of Finance 49(1):3–37.

Prendergast, Canice. 1999. “The Provision of Incentives in Firms.” Journal of EconomicLiterature 37(1):7–63.

Qian, Jun, Philip E. Strahan and Zhishu Yang. 2011. “The Impact of Incentives and Com-munication costs on Information Production: Evidence from Bank Lending.” WorkingPaper, Boston College .

Scheier, M., C. Carver and M. Bridges. 1994. “Distinguishing Optimism from Neuroticism(and Trait-Anxiety, Self-Mastery, and Self-Esteem): A Re-Evaluation of the Life Orienta-tion Test.” Journal of Personality and Social Psychology 67:1063–1078.

Shapiro, Jesse. 2005. “Is there a daily discount rate? Evidence from the food stamp nutritioncycle.” Journal of Public Economics 89(2):303–325.

Stein, Jeremy C. 2002. “Information Production and Capital Allocation: Decentralizedversus Hierarchical Firms.” The Journal of Finance 57(5):1891–1921.

39

Tables and Figures

Table ILoan Officer Summary Statistics

This table reports summary statistics for loan officer demographics and personality traits. Panel A, columns[1] to [4] report demographic characteristics for all participants of the experiment. Panel A, columns [5]to [8] report summary statistics for the same demographic characteristics for the loan officer populationof a large public sector bank as a basis for comparison. Male is a dummy variable equal to one if theparticipant is male. Age is the loan officer’s age. Experience is the number of years the loan officer has beenemployed with the bank. Seniority is the loan officer’s seniority rank, ranging from 1 (lowest) to 5 (highest).Education is a dummy equal to one if a loan officer has a master’s degree or equivalent qualification. Privatesector banker is a dummy equal to one if a loan officer is employed by a private sector bank. Panel B reportssummary statistics for the tests of attitudes and personality characteristics completed by participants ofthe experiment. Impatience is a dummy equal to one if a loan officer’s monthly discount rate is in the topdecile of the sample distribution. Risk averse is a dummy equal to one if a participant states that she neverplays the lottery. Optimism is the LOT-R test measure of optimism (Scheier, Carver and Bridges [1994]).Conscientiousness is the ‘Big Five’ (BFI) personality test measure of conscientiousness (John, Donahue andKentle [1991]). Confidence is a dummy equal to one if a loan officer ranks herself in the top decile of theperformance distribution. Overconfidence is a dummy equal to one if a loan officer incorrectly ranks herselfin the top decile of the performance distribution, based on the realized outcome of all lending decisionsmade in the experiment.

Panel A: Demographics

Experiment participants [N=209] Bank sample [N=3,111]

N Mean Median StdDev N Mean Median StdDev

(1) (2) (3) (4) (5) (6) (7) (8)

Male 206 0.90 1.00 0.30 3,111 0.9 1.00 0.30

Age 206 37.60 35.00 10.94 3,111 37.9 35.00 12.0

Experience [Years] 206 12.76 10.00 11.30 3,111 13.90 11.00 13.00

Seniority [1 (Lowest) - 5 (Highest)] 206 1.94 2.00 1.00 3,111 1.60 2.00 0.75

Education [Master’s degree] 200 0.33 0.00 0.47 N/A N/A N/A N/A

Private sector banker 206 0.26 0.00 0.43 3,111 0.00 0.00 0.00

Panel B: Personality traits

Experiment participants

N Mean Median StdDev p10 p25 p75 p90

(1) (2) (3) (4) (5) (6) (7) (8)

Impatient 74 0.11 0.00 0.33 0.00 0.00 0.00 1.00

Risk averse 172 0.75 1.00 0.43 0.00 0.50 1.00 1.00

Optimism [LOT-R] 64 2.37 2.33 0.46 1.83 2.17 2.67 3.00

Conscientiousness [BFI] 72 3.81 3.89 0.47 3.11 3.50 4.17 4.44

Confidence 71 0.73 0.78 0.20 0.50 0.60 0.85 0.99

Overconfidence 69 0.19 0.00 0.39 0.00 0.00 0.00 1.00

40

Table

IIS

um

mary

of

Ince

nti

ve

Tre

atm

ents

Th

ista

ble

sum

mar

izes

the

exp

erim

enta

lin

centi

ve

sch

emes

.E

ach

ince

nti

vesc

hem

eco

nsi

sts

of

aco

nd

itio

nal

pay

men

tw

Pfo

rap

pro

vin

ga

loan

that

per

form

s,a

con

dit

ion

al

pay

men

tw

Dfo

rap

pro

vin

ga

loan

that

def

au

lts

an

dan

ou

tsid

ep

aym

ent

wfo

rd

ecli

nin

ga

loan

,in

wh

ich

case

the

outc

ome

of

the

loan

isn

ot

ob

serv

ed.

All

ince

nti

ves

refe

rto

con

dit

ion

alp

ayoff

sfo

ran

ind

ivid

ual

len

din

gd

ecis

ion

.T

he

ince

nti

vep

aym

ents

[wP

,w

D,

w]

for

each

ince

nti

vesc

hem

eare

rep

ort

edin

colu

mn

[1].

Colu

mn

[2]

rep

ort

sth

enu

mb

erof

obse

rvat

ion

sby

ince

nti

vesc

hem

e,an

dco

lum

ns

[3]

to[8

]re

port

sam

ple

size

sfo

rth

esu

bse

tof

ob

serv

ati

on

sw

ith

the

‘cos

tly

info

rmat

ion

’,‘d

efer

red

com

pen

sati

on

’an

d‘lim

ited

liab

ilit

y’

featu

res.

Un

der

the

‘cost

lyin

form

ati

on

’fe

atu

re,

part

icip

ants

are

char

ged

cred

its

tore

vie

wad

dit

ion

alse

ctio

ns

ofth

elo

an

file

ap

pli

cati

on

.F

or

loan

sev

alu

ate

du

nd

erth

e‘d

efer

red

com

pen

sati

on’

feat

ure

,lo

anoffi

cers

rece

ive

allea

rned

ince

nti

vep

aym

ents

wit

ha

thre

em

onth

lag,u

nd

erth

e‘l

imit

edli

ab

ilit

y’fe

atu

re,lo

an

offi

cers

are

given

anen

dow

men

t,w

hic

hth

eyca

nlo

seby

makin

gb

ad

len

din

gd

ecis

ion

s.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Incenti

ve

Incenti

ve

Ob

servati

on

sC

ost

lyD

efe

rred

Lim

ited

Treatm

ent

Paym

ents

[am

ou

nt

inR

s]In

form

ati

on

Com

pen

sati

on

Lia

bil

ity

[Per

form|D

efau

lt|D

eclin

e]N

oY

esN

oY

esN

oY

es

AL

ow

-Pow

ered

[Base

lin

e][2

0,

0,

10]

7,4

20

3,7

82

3,6

38

6,5

68

852

N/A

7,4

20

BH

igh

-Pow

ered

[50,

-100,

0]

2,9

46

654

2,2

92

2,4

96

450

978

1,9

68

CO

rigin

ati

on

Bonus

[20,

20,

0]

2,5

48

762

1,7

86

1,6

32

916

N/A

2,5

48

41

Table

III

Loan

Evalu

ati

on

Sum

mary

Sta

tist

ics

Th

ista

ble

rep

orts

sum

mar

yst

atis

tics

for

the

len

din

gd

ecis

ion

sm

ad

ein

the

exp

erim

ent

by

ince

nti

vetr

eatm

ent.

Th

eta

ble

rep

ort

su

nco

nd

itio

nal

mea

ns

and

stan

dar

dd

evia

tion

sfo

rlo

anev

alu

ati

on

sm

ad

eu

nd

erb

ase

lin

e,h

igh

-pow

ered

an

dori

gin

ati

on

ince

nti

ves.

Len

din

gd

ecis

ion

sm

ad

eu

nd

erth

e‘d

efer

red

ince

nti

ves’

and

‘lim

ited

liab

ilty

’tr

eatm

ent

con

dit

on

sare

excl

ud

ed.

Th

efi

rst

two

colu

mn

sof

the

tab

lere

port

sum

mary

stat

isti

csfo

rsc

reen

ing

effor

t,S

ecti

on

sre

view

edis

the

nu

mb

erof

loan

file

sect

ion

sre

vie

wed

.A

mou

nt

spen

ton

info

rmati

on

isth

enu

mb

erof

info

rmat

ion

cred

its

spen

tby

loan

office

rsfo

rlo

an

evalu

ati

on

sm

ad

eu

nd

erth

e‘c

ost

lyin

form

ati

on

’co

nd

itio

n,

inw

hic

hp

art

icip

ants

wer

ech

arge

dto

acce

ssin

form

atio

nb

eyon

dth

eb

asic

ap

pli

cant

det

ail

s.R

isk

rati

ng

isth

ein

tern

al

risk

-rati

ng

ass

ign

edto

loan

sev

alu

ate

du

nd

era

give

ntr

eatm

ent

cond

itio

n,

wit

hh

igh

erra

tin

gsin

dic

ati

ng

bet

ter

loan

qu

ali

ty.

Appro

ved

isth

esh

are

of

loan

sap

pro

ved

.P

rofi

tis

the

pro

fit

per

scre

ened

loan

.In

colu

mn

s[7

]to

[10]

,w

ere

por

tth

esh

are

of

corr

ect

len

din

gd

ecis

ion

sby

ince

nti

vetr

eatm

ent

for

the

sam

ple

as

aw

hole

,an

dth

esu

bsa

mp

les

ofp

erfo

rmin

g,n

on-p

erfo

rmin

gan

dex

-ante

dec

lin

edlo

an

ap

pli

cati

on

s,re

spec

tivel

y.A

corr

ect

len

din

gd

ecis

ion

isd

efined

as

app

rovin

ga

loan

app

lica

tion

that

ex-p

ost

per

form

ed,

or

turn

ing

dow

na

loan

ap

pli

cati

on

that

was

eith

ersc

reen

dou

tby

the

len

der

or

def

au

lted

.

Eff

ort

Ris

krati

ng

Ap

proved

Profi

tp

er

loan

Evalu

ati

on

sC

orrect

Sec

tion

sA

mou

nt

spen

t%

scre

ened

ap

pro

ved

Sam

ple

Per

form

ing

Non

-D

eclin

ed

revie

wed

on

info

rmati

on

per

form

ing

by

ban

k

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

Sam

ple

5.0

72.3

171.6

90.7

4542.9

1699.0

20.6

40.8

00.2

80.4

6

(2.3

6)

(3.7

6)

(15.7

7)

(0.4

4)

(1870.0

8)

(2096.1

4)

(0.4

8)

(0.4

0)

(0.4

5)

(0.5

0)

Base

lin

e5.2

52.0

170.8

20.7

2529.9

8693.0

40.6

40.7

80.2

80.5

1

(2.4

1)

(3.5

7)

(15.7

0)

(0.4

5)

(1864.0

7)

(2104.9

9)

(0.4

8)

(0.4

1)

(0.4

5)

(0.5

0)

Hig

h-p

ow

ered

5.0

13.1

974.0

00.6

9576.1

0801.4

50.6

40.7

50.3

90.4

8

(2.3

3)

(4.4

2)

(15.6

7)

(0.4

6)

(1732.7

0)

(1999.2

2)

(0.4

8)

(0.4

3)

(0.4

9)

(0.5

0)

Ori

gin

ati

on

4.5

92.5

272.4

50.8

4554.8

3648.3

70.6

50.8

70.1

90.2

9

(2.1

8)

(3.6

1)

(15.8

3)

(0.3

7)

(1986.0

7)

(2132.8

8)

(0.4

8)

(0.3

3)

(0.3

9)

(0.4

5)

42

Table IVThe Effect of Incentives on Effort

This table estimates the effect of performance pay on screening effort. Each column reportsresults from a separate regression. The omitted category in all regressions is the low-poweredbaseline incentive. The dependent variable in column [1] and [2] is the number of loan filesections reviewed, the dependent variable in columns [3] and [4] is the number of loan filesections reviewed when loan officers were required to pay for additional information. Theregressions in columns [1] and [2] include the entire sample, while columns [3] and [4] restrictthe sample to loan evaluations made under the ‘costly information’ condition. All regressionsinclude a lab fixed effect, randomization stratum and week fixed effects, as well as dummiesto control for treatment conditions not reported in this table. Loan officer controls includeage, seniority, rank, education, and indicators for branch manager and business experience.Standard errors, in parentheses, are clustered at the loan officer × session level, the same levelof observation at which the treatment is assigned. * p<0.10 ** p<0.05 *** p<0.01.

Free information Costly information

Loan file Loan file

sections reviewed sections reviewed

(1) (2) (3) (4)

Baseline, omitted

High-powered 0.434* 0.400*** 1.225*** 0.794***

(0.23) (0.14) (0.42) (0.25)

Origination bonus 0.083 0.005 -0.147 -0.156

(0.22) (0.14) (0.40) (0.21)

Loan fixed effects No Yes No Yes

Loan officer fixed effects No Yes No Yes

Loan officer controls Yes No Yes No

Number of observations 14,405 14,675 8,520 8,688

R-squared, adjusted 0.452 0.698 0.266 0.694

43

Table VThe Effect of Incentives on Risk-Assessment

This table reports the effect of performance pay on loan officers’ subjective assessment ofcredit risk. Each column shows results from a separate regression. The omitted category inall columns is the low-powered baseline incentive. The dependent variable in regressions [1]and [2] is the overall risk rating, normalized to have mean zero and standard deviation 1. Thedependent variable in columns [3] and [4] is the normalized sub-rating for all categories thatpertain to the personal risk of a potential applicant. In columns [5] and [6] the dependentvariable is the normalized sub-rating for all rating categories that pertain to the business,management and financial risk of a loan applicant. All regressions include a lab fixed effect,randomization stratum and week fixed effects, as well as dummies to control for treatmentconditions not reported in this table. Standard errors, in parentheses, are clustered at theloan officer × session level. * p<0.10 ** p<0.05 *** p<0.01.

Internal rating

Overall rating Personal and Business and

management risk financial risk

(1) (2) (3) (4) (5) (6)

Baseline, omitted

High-powered 0.029 0.007 0.012 -0.001 0.054 0.020

(0.09) (0.04) (0.09) (0.04) (0.09) (0.04)

Origination bonus 0.145* 0.006 0.132* -0.015 0.157** 0.021

(0.08) (0.04) (0.08) (0.04) (0.08) (0.04)

Loan fixed effects No Yes No Yes No Yes

Loan officer fixed effects No Yes No Yes No Yes

Loan officer controls Yes No Yes No Yes No

Number of observations 14,405 14,675 14,405 14,675 14,405 14,675

R-squared, adjusted 0.147 0.615 0.137 0.619 0.156 0.600

44

Table

VI

The

Eff

ect

of

Ince

nti

ves

on

Ris

k-T

akin

g

Th

ista

ble

esti

mat

esth

eeff

ect

ofp

erfo

rman

cep

ayon

risk

-takin

g.

Each

colu

mn

rep

ort

sre

sult

sfr

om

ase

para

tere

gre

ssio

n.

Th

eom

itte

dca

tegory

inal

lre

gres

sion

sis

the

low

-pow

ered

bas

elin

ein

centi

ve.

Th

ed

epen

den

tva

riab

lein

colu

mn

s[1

]th

rou

gh

[6]

isa

mea

sure

of

the

per

ceiv

edqu

ality

of

the

loan

:th

eav

erag

ein

tern

alra

tin

gof

each

loan

rep

ort

edby

all

loan

offi

cers

un

der

the

base

lin

etr

eatm

ent.

To

cap

ture

the

deg

ree

of

ex-a

nte

un

cert

ain

tyab

out

the

qu

alit

yof

alo

an,

colu

mn

s[7

]to

[12]

rep

eat

the

exer

cise

usi

ng

the

coeffi

cien

tof

vari

ati

on

of

inte

rnal

rati

ng

ass

ign

edto

agiv

enlo

an

un

der

the

bas

elin

etr

eatm

ent

asth

edep

end

ent

vari

able

.T

he

inte

rnal

rati

ng

isn

orm

ali

zed

toh

ave

mea

nze

roan

dst

an

dard

dev

iati

on

1,

hen

ceeff

ect

size

sar

est

and

ard

dev

iati

ons.

All

regr

essi

ons

incl

ude

ala

bfi

xed

effec

t,ra

nd

om

izati

on

stra

tum

an

dw

eek

fixed

effec

ts,

as

wel

las

du

mm

ies

toco

ntr

ol

for

trea

tmen

tco

nd

itio

ns

not

rep

orte

din

this

tab

le.

Sta

nd

ard

erro

rs,

inp

are

nth

eses

,are

clu

ster

edat

the

loan

offi

cer×

sess

ion

leve

l.*

p<

0.1

0**

p<

0.0

5**

*p<

0.01

.

Perceiv

ed

qu

ali

tyof

ap

proved

loan

sP

erceiv

ed

loan

qu

ali

tyof

ap

proved

loan

s

[Mea

nra

tin

g]a

[Coeffi

cien

tof

vari

ati

on

]b

Over

all

rati

ng

Per

son

al

an

dB

usi

nes

san

dO

ver

all

rati

ng

Per

son

al

an

dB

usi

nes

san

d

man

agem

ent

risk

fin

an

cial

risk

man

agem

ent

risk

fin

an

cial

risk

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Base

lin

e,om

itte

d

Hig

h-p

ow

ered

0.0

58

-0.0

33

0.0

41

-0.0

41

0.0

91

-0.0

09

-0.0

15***

-0.0

15***

-0.0

18***

-0.0

18***

-0.0

13**

-0.0

13**

(0.1

0)

(0.0

3)

(0.0

99)

(0.0

3)

(0.1

0)

(0.0

3)

(0.0

1)

(0.0

1)

(0.0

1)

(0.0

1)

(0.0

1)

(0.0

1)

Ori

gin

ati

on

bonu

s0.0

80

-0.0

17

0.0

76

-0.0

26

0.0

95

-0.0

03

-0.0

08*

-0.0

07

-0.0

07

-0.0

06

-0.0

10**

-0.0

09*

(0.0

8)

(0.0

4)

(0.0

82)

(0.0

4)

(0.0

8)

(0.0

4)

(0.0

0)

(0.0

0)

(0.0

0)

(0.0

0)

(0.0

0)

(0.0

0)

Loan

fixed

effec

tsN

oN

oN

oN

oN

oN

oN

oN

oN

oN

oN

oN

o

Loan

offi

cer

effec

tsN

oY

esN

oY

esN

oY

esN

oY

esN

oY

esN

oY

es

Loan

offi

cer

contr

ols

Yes

No

Yes

No

Yes

No

Yes

No

Yes

No

Yes

No

Nu

mb

erof

ob

serv

ati

on

s10,7

15

10,4

02

10,7

15

10,9

48

10,7

15

10,9

48

9,3

49

9,5

55

9,3

49

9,5

55

9,3

49

9,5

55

R-s

qu

are

d,

ad

just

ed0.1

96

0.8

02

0.0

70.7

73

0.2

06

0.7

88

0.0

52

0.0

53

0.0

54

0.0

56

0.0

54

0.0

55

[a]

Mea

nra

tin

gass

ign

edto

loan

ap

plica

tionl

by

all

loan

offi

cers

evalu

ati

ng

the

loan

un

der

the

base

lin

etr

eatm

ent.

[b]

Coeffi

cien

tof

vari

ati

on

of

rati

ngs

ass

iged

tolo

an

ap

plica

tionl

by

all

loan

offi

cers

revie

win

gth

elo

an

un

der

the

base

lin

etr

eatm

ent.

45

Table

VII

Ince

nti

ves,

Lendin

gD

eci

sions

and

Pro

fit

Th

ista

ble

rep

orts

the

effec

tof

per

form

ance

pay

on

loan

ap

pro

vals

inP

an

elA

(colu

mn

s[1

]to

[8])

and

the

pro

fita

bil

ity

of

len

din

gin

Pan

elB

(colu

mn

s[9

]to

[12]

).E

ach

colu

mn

rep

orts

resu

lts

from

ase

para

tere

gre

ssio

n.

Th

eom

itte

dtr

eatm

ent

cate

gory

isth

elo

w-p

ower

edb

ase

lin

ein

centi

ve.

Th

ed

epen

den

tva

riab

lein

colu

mn

s[1

]to

[8]

isa

du

mm

yva

riab

leeq

ual

toon

efo

rlo

an

sap

pro

ved

by

an

exp

erim

enta

lp

art

icip

ant

an

dze

rooth

erw

ise.

Th

ees

tim

ate

sin

colu

mn

s[1

]an

d[2

]ar

eb

ased

onth

efu

llsa

mp

le.

Est

imate

sin

colu

mns

[3]

an

d[4

]are

base

don

the

sam

ple

of

per

form

ing

loan

s,es

tim

ate

sin

colu

mn

s[5

]an

d[6

]ar

eb

ased

onth

esa

mp

leof

non

-per

form

ing

loan

s,an

des

tim

ate

sin

colu

mn

s[7

]an

d[8

]are

base

don

the

sam

ple

of

loan

sth

at

wer

ein

itia

lly

dec

lin

edby

the

Len

der

.C

olu

mn

s[9

]to

[12]

rep

ort

trea

tmen

tes

tim

ate

sof

ince

nti

ves

on

pro

fit

per

ap

pro

ved

loan

an

dp

rofi

tp

ersc

reen

edlo

an

inU

S$.

All

regr

essi

ons

incl

ud

ea

lab

fixed

effec

t,ra

nd

omiz

atio

nst

ratu

man

dw

eek

fixed

effec

ts,

as

wel

las

du

mm

ies

toco

ntr

ol

for

trea

tmen

tco

nd

itio

ns

not

rep

ort

edin

this

tab

le.

Loa

noffi

cer

contr

ols

incl

ud

eag

e,se

nio

rity

,ra

nk,

edu

cati

on

,an

din

dic

ato

rsfo

rb

ran

chm

an

ager

an

db

usi

nes

sex

per

ien

ce.

Sta

nd

ard

erro

rs,

inp

aren

thes

es,

are

clu

ster

edat

the

loan

office

r×

sess

ion

level

.*

p<

0.1

0**

p<

0.0

5***

p<

0.0

1.

Pan

el

A:

Ap

proved

Pan

el

B:

Profi

t

Tota

lP

erfo

rmin

gN

on

-per

form

ing

Dec

lin

edby

ban

kp

erap

pro

ved

loan

per

scre

ened

loan

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Base

lin

e,om

itte

d

Hig

h-p

ow

ered

-0.0

36*

-0.0

04

-0.0

10

0.0

15

-0.1

10**

-0.0

63

-0.0

42

-0.0

14

148.9

86*

175.9

07**

84.9

00

114.9

30*

(0.0

2)

(0.0

2)

(0.0

3)

(0.0

3)

(0.0

6)

(0.0

6)

(0.0

6)

(0.0

6)

(85.0

1)

(86.8

1)

(62.5

1)

(63.7

6)

Ori

gin

ati

on

bonu

s0.0

83***

0.0

79***

0.0

87***

0.0

68***

0.0

48

0.0

82*

0.0

98*

0.1

02*

29.4

89

-4.1

82

80.1

93

56.5

00

(0.0

2)

(0.0

2)

(0.0

2)

(0.0

2)

(0.0

5)

(0.0

5)

(0.0

6)

(0.0

5)

(78.0

4)

(79.0

2)

(60.9

6)

(61.2

1)

Loan

fixed

effec

tsN

oY

esN

oY

esN

oY

esN

oY

esN

oN

oN

oN

o

Loan

offi

cer

fixed

effec

tsN

oY

esN

oY

esN

oY

esN

oY

esN

oY

esN

oY

es

Loan

offi

cer

contr

ols

Yes

No

Yes

No

Yes

No

Yes

No

Yes

No

Yes

No

Nu

mb

erof

ob

serv

ati

on

s14,4

05

14,6

75

9,3

98

9,5

75

2,7

30

2,7

78

2,2

77

2,3

22

9,2

42

9,4

35

11,8

53

12,0

74

R-s

qu

are

d,

ad

just

ed0.0

19

0.1

57

0.0

17

0.1

24

0.0

26

0.1

40

0.0

48

0.1

96

0.0

09

0.0

10

0.0

07

0.0

16

46

Table

VII

ID

efe

rred

Com

pensa

tion

Th

ista

ble

rep

orts

trea

tmen

teff

ects

ofd

efer

rin

gp

erfo

rman

cepay

by

thre

em

onth

s.E

ach

colu

mn

rep

ort

sre

sult

sfr

om

ase

para

tere

gre

ssio

n.

Th

eom

itte

dtr

eatm

ent

cate

gory

isth

elo

w-p

ower

edb

asel

ine

con

dit

ion

.T

he

dep

end

ent

vari

ab

lein

colu

mn

s[1

]an

d[2

]is

the

nu

mb

erof

loan

file

sect

ion

sre

vie

wed

for

each

eval

uat

edlo

an.

Th

ed

epen

den

tva

riab

lein

colu

mn

s[3

]an

d[4

]is

the

am

ou

nt

spen

ton

revie

win

gad

dit

ion

al

info

rmati

on

un

der

the

“co

stly

info

rmat

ion

”co

nd

itio

n.

Inco

lum

ns

[5]

and

[6]

we

con

sid

erth

eeff

ect

of

def

erre

dco

mp

ensa

tion

on

risk

-takin

g.

The

dep

end

ent

vari

ab

leis

the

mea

nan

dco

effici

ent

ofva

riati

onof

inte

rnal

rati

ngs

assi

gned

toea

chlo

an

un

der

the

base

lin

efo

rlo

an

sap

pro

ved

by

part

icip

ants

inth

eex

per

imen

tas

the

ou

tcom

eof

inte

rest

,w

ith

the

sam

ple

rest

rict

edto

loan

sth

elo

an

offi

cer

ap

pro

ves.

Th

ed

epen

den

tva

riab

lein

colu

mn

s[7

]an

d[8

]is

ad

um

my

equ

al

toon

eif

alo

an

eval

uat

edin

the

exp

erim

ent

was

app

rove

dan

dze

rooth

erw

ise.

Th

ed

epen

den

tva

riab

les

inco

lum

ns

[9]

an

d[1

0]

rep

ort

trea

tmen

tes

tim

ate

sof

mon

etary

ince

nti

ves

onp

rofi

tp

erap

pro

ved

loan

and

pro

fit

per

scre

ened

loan

,d

enom

inate

din

US

$.

All

regre

ssio

ns

incl

ud

ea

lab

fixed

effec

t,ra

nd

om

izati

on

stra

tum

and

wee

kfi

xed

effec

ts,

asw

ell

asd

um

mie

sto

contr

ol

for

trea

tmen

tco

nd

itio

ns

not

rep

ort

edin

this

tab

le.

Loan

offi

cer

contr

ols

incl

ud

eage,

sen

iori

ty,

ran

k,

edu

cati

on,

and

ind

icat

ors

for

bra

nch

man

ager

an

db

usi

nes

sex

per

ien

ce.

Tes

tst

ati

stic

sat

the

foot

of

the

tab

lere

fer

tot-

test

sfo

rth

eeq

uali

tyof

coeffi

cien

tsb

etw

een

the

def

erre

dan

dn

on-d

efer

red

trea

tmen

td

um

mie

s.S

tan

dard

erro

rs,

inpare

nth

eses

,are

clu

ster

edat

the

loan

offi

cer×

sess

ion

leve

l.*

p<

0.10

**p<

0.05

***

p<

0.01

. Pan

el

A:

Screen

ing

eff

ort

Pan

el

B:

Ris

k-t

akin

gP

an

el

C:

Len

din

gan

dp

rofi

t

Loan

file

Am

ou

nt

spen

tA

ver

age

Inte

rnal

Rati

ng

Ap

pro

ved

Pro

fit

per

loan

sect

ion

sre

vie

wed

on

info

rmati

on

Mea

ncv

ap

pro

ved

scre

ened

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

Base

lin

e,om

itte

d

Base

lin

e,d

efer

red

-0.2

05

-0.1

46*

-0.5

38

-0.2

48

0.1

00***

-0.0

13***

0.0

00

0.0

25

-144.1

37*

-133.3

37*

(0.1

40)

(0.0

77)

(0.3

5)

(0.2

0)

(0.0

4)

(0.0

0)

(0.0

2)

(0.0

2)

(73.7

9)

(70.0

3)

Hig

h-p

ow

ered

0.3

45**

0.2

48**

1.2

25***

0.7

94***

0.0

81**

-0.0

11**

-0.0

48**

-0.0

60***

39.6

56

78.6

82

(0.1

59)

(0.1

00)

(0.4

2)

(0.2

5)

(0.0

3)

(0.0

1)

(0.0

2)

(0.0

2)

(71.5

0)

(65.6

8)

Hig

h-p

ow

ered

,d

efer

red

-0.2

69

0.0

08

-0.4

54

0.0

34

0.0

48

-0.0

12*

-0.0

21

-0.0

17

-60.3

65

-52.4

66

(0.2

09)

(0.1

21)

(0.5

0)

(0.2

9)

(0.0

4)

(0.0

1)

(0.0

3)

(0.0

3)

(105.1

3)

(100.1

0)

Ori

gin

ati

on

bonu

s-0

.165

-0.1

25

-0.1

47

-0.1

56

0.1

84***

-0.0

16***

0.1

10***

0.0

93***

-165.9

72**

-65.1

90

(0.1

54)

(0.0

81)

(0.4

0)

(0.2

1)

(0.0

4)

(0.0

0)

(0.0

2)

(0.0

2)

(74.7

3)

(69.8

4)

Ori

gin

ati

on

bonu

s,d

efer

red

-0.0

76

-0.1

45

-0.2

07

-0.3

87*

0.1

85***

-0.0

08*

0.0

79***

0.0

90***

-55.1

38

-15.0

03

(0.1

38)

(0.0

90)

(0.3

7)

(0.2

3)

(0.0

4)

(0.0

0)

(0.0

2)

(0.0

2)

(65.1

1)

(62.4

8)

Loan

effec

tsN

oY

esN

oY

esN

oN

oN

oY

esN

oN

o

Loan

offi

cer

effec

tsN

oY

esN

oY

esY

esY

esN

oY

esY

esY

es

Loan

offi

cer

contr

ols

Yes

No

Yes

No

No

No

Yes

No

No

No

Test:

immediate=deferred

Base

lin

e[0

.14]

[006]

[0.1

2]

[0.2

1]

[0.0

1]

[0.0

1]

[1.0

0]

[0.1

9]

[0.0

5]

[0.0

6]

Hig

h-p

ow

ered

[0.0

1]

[0.0

0]

[0.0

0]

[0.0

2]

[0.3

8]

[0.9

0]

[0.3

3]

[0.1

0]

[0.3

5]

[0.1

9]

Ori

gin

ati

on

bonu

s[0

.56]

[0.8

1]

[0.8

8]

[0.2

8]

[0.9

7]

[0.0

7]

[0.0

8]

[0.8

8]

[0.1

0]

[0.4

4]

Nu

mb

erof

ob

serv

ati

on

s8,5

20

8,6

88

8,5

20

8,6

88

8,0

90

7,2

63

8,5

20

8,5

20

5,6

19

7,7

41

R-s

qu

are

d,

ad

just

ed0.2

77

0.6

93

0.2

66

0.6

94

0.0

40

0.0

39

0.0

47

0.1

55

0.6

47

0.4

96

47

Table

IXR

ela

xin

gL

imit

ed

Lia

bil

ity

Th

ista

ble

rep

orts

the

effec

tof

rela

xin

glo

anoffi

cers

’li

mit

edli

ab

ilit

yco

nst

rain

t.E

ach

colu

mn

rep

ort

sre

sult

sfr

om

ase

para

tere

gre

ssio

n,

the

om

itte

dca

tego

ryin

each

regr

essi

onis

the

low

-pow

ered

base

lin

etr

eatm

ent.

Pan

elA

(colu

mn

s[1

]to

[4])

rep

ort

trea

tmen

teff

ects

on

scre

enin

geff

ort

,P

an

elB

(col

um

ns

[5]

and

[6])

rep

ort

trea

tmen

teff

ects

onri

sk-t

akin

gan

dP

an

elC

(colu

mn

s[6

]to

[8])

rep

ort

trea

tmen

teff

ects

on

loan

ap

pro

vals

an

dp

rofi

tp

erap

pro

ved

loan

.T

he

dep

end

ent

vari

able

inco

lum

ns

[7]to

[8]is

ad

um

my

equ

alto

1if

alo

an

evalu

ate

din

the

exp

erim

ent

was

ap

pro

ved

an

dze

rooth

erw

ise.

Th

ed

epen

den

tva

riab

lein

colu

mn

s[9

]an

d[1

0]are

the

ban

k’s

pro

fit

per

ap

pro

ved

loan

an

dth

eb

an

k’s

pro

fit

per

scre

ened

loan

,re

spec

tive

ly,

den

om

inate

din

US

$.A

llre

gres

sion

sin

clu

de

ala

bfi

xed

effec

t,ra

nd

om

izati

on

stra

tum

an

dw

eek

fixed

effec

ts,

as

wel

las

du

mm

ies

toco

ntr

ol

for

trea

tmen

tco

nd

itio

ns

not

rep

orte

din

this

tab

le.

Loa

noffi

cer

contr

ols

incl

ud

eage,

sen

iori

ty,

ran

k,

edu

cati

on

,an

din

dic

ato

rsfo

rb

ran

chm

an

ager

an

db

usi

nes

sex

per

ien

ce.

Tes

tst

atis

tics

atth

efo

otof

the

tab

lere

fer

tot-

test

sfo

rth

eeq

uali

tyof

coeffi

cien

tsb

etw

een

the

hig

h-p

ow

ered

trea

tmen

td

um

mie

sw

hen

lim

ited

liab

ilit

yis

pre

sent

vs.

rela

xed

.S

tan

dar

der

rors

,in

par

enth

eses

,are

clu

ster

edat

the

loan

offi

cer×

sess

ion

level

.*

p<

0.1

0**

p<

0.0

5***

p<

0.0

1.

Pan

el

A:

Screen

ing

eff

ort

Pan

el

B:

Ris

k-t

akin

gP

an

el

C:

Len

din

gan

dp

rofi

t

Loan

file

Info

rmati

on

Inte

rnal

rati

ng

[base

lin

e]A

pp

roved

Pro

fit

per

loan

sect

ion

sre

vie

wed

cred

its

spen

tM

ean

cvap

pro

ved

scre

ened

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

Base

lin

e,om

itte

d

[cre

dit

]

Hig

h-p

ow

ered

0.3

45**

0.2

48**

1.2

25***

0.7

94***

0.0

81**

-0.0

11**

-0.0

48**

-0.0

60***

39.6

56

78.6

82

[cre

dit

](0

.16)

(0.1

0)

(0.4

2)

(0.2

5)

(0.0

3)

(0.0

1)

(0.0

2)

(0.0

2)

(71.5

0)

(65.6

8)

Hig

h-p

ow

ered

0.5

55***

0.3

72***

1.9

00***

1.2

60***

-0.0

57*

0.0

06

-0.0

77***

-0.0

74***

49.9

00

22.9

40

[cre

dit

+en

dow

men

t](0

.16)

(0.0

8)

(0.4

4)

(0.2

2)

(0.0

3)

(0.0

1)

(0.0

2)

(0.0

2)

(80.0

4)

(69.5

0)

Loan

effec

tsN

oY

esN

oY

esN

oN

oN

oY

esN

oN

o

Loan

offi

cer

effec

tsN

oY

esN

oY

esY

esY

esN

oY

esY

esY

es

Loan

offi

cer

contr

ols

Yes

No

Yes

No

No

No

Yes

No

No

No

Test:

High-P

owered

noen

dowmen

t=

High-powered

withen

dowmen

t:[0

.34]

[0.3

4]

[0.2

5]

[0.1

7]

[0.0

0]

[0.0

0]

[0.3

3]

[0.6

3]

[0.9

2]

[0.5

3]

Nu

mb

erof

ob

serv

ati

on

s8,5

20

8,6

88

8,5

20

8,6

88

6,1

00

5,4

63

8,5

20

8,6

88

5,6

94

7,2

22

R-s

qu

are

d,

ad

just

ed0.2

77

0.6

93

0.2

66

0.6

94

0.0

40

0.0

39

0.0

47

0.1

55

0.6

47

0.4

96

48

Table XHeterogeneity in the Response to Incentives

This table examines the interaction between incentive schemes and loan officer personality traits. In each panel,the first two columns report the main effect of the personality characteristic indicated in the panel heading,the second two columns report interactions. All personality traits are as defined in Table I. Further details onthe measurement of personality traits are available in the Internet Appendix. All regressions control for loanapplication fixed effects, loan officer age, rank, gender, education, a lab fixed effect, randomization stratumand week fixed effects. Regressions in Panel B additionally control for all measured personality traits andnon-reported categories of the BFI personality test. Standard errors, in parentheses, are clustered at the loanofficer × session level. * p<0.10 ** p<0.05 *** p<0.01.

Screening Effort

Sections reviewed Information credits spent

Main Effect Interaction Main Effect Interaction

(1) (2) (3) (4) (5) (6) (7) (8)

Panel A: Career Concerns

A.1 Age -0.05*** (0.01) -0.12***

High-powered 0.39 (0.84) 0.00 (0.02) 7.43* (4.39) 0.02 (0.03)

Origination 0.87 (0.86) -0.02 (0.02) 9.14** (4.49) -0.07 (0.05)

R-squared, N 0.500 6,102 0.42 3,828

A.2 Private Banker 0.39** (0.17) 1.50*** (0.45)

High-powered 0.78** (0.33) -0.69 (0.43) 1.59* (0.85) -0.71 (0.79)

Origination 0.31 (0.31) -0.46 (0.42) 0.72 (0.94) -1.76** (0.75)

R-squared, N 0.456 14,405 0.284 8,520

Panel B: Personality Traits

B.1 Impatience -0.54 (0.55) -0.35 (1.03)

High-powered 0.29 (0.28) 2.94*** (0.75) 1.26** (0.62) 2.99 (2.27)

Origination 0.15 (0.30) 1.33 (0.92) -0.23 (0.70) 2.40 (1.96)

R-squared, N 0.503 6,102 0.436 3,828

B.2 Risk-aversion 1.53*** (0.32) 1.36** (0.57)

High-powered -0.87 (0.86) 1.33 (0.89) 3.63** (1.74) -1.66 (1.51)

Origination 1.28 (0.88) -1.21 (0.93) -0.08 (1.37) 0.96 (1.12)

R-squared, N 0.504 6,102 0.421 3,828

B.3 Optimism 0.44 (0.31) -0.19 (0.67)

High-powered 0.55* (0.29) -0.79 (0.56) 1.43** (0.64) -2.27** (1.10)

Origination 0.33 (0.30) -0.98 (0.92) 0.06 (0.73) -4.08*** (1.31)

R-squared, N 0.500 6,102 0.424 3,828

B.4 Conscientiousness -0.37 (0.29) -0.93 (0.59)

High-powered -5.93*** (1.70) 1.77*** (0.46) -2.88 (4.76) 1.55 (0.95)

Origination -6.84*** (2.15) 1.93*** (0.59) -0.28 (4.69) 0.41 (0.91)

R-squared, N 0.507 6,102 0.421 3,828

B.5 Overconfidence -1.00** (0.45) -1.12* (0.67)

High-powered 0.54* (0.28) -0.21 (0.86) 1.22* (0.68) 1.26 (1.35)

Origination 0.05 (0.29) 1.14 (0.84) -0.26 (0.82) 1.25 (1.04)

R-squared, N 0.500 6,102 0.427 3,82849

01

23

45

Den

sity

.3 .4 .5 .6 .7 .8

% of Lending Decisions Correct [Baseline]

Figure 1: Loan Officer Performance. This figure shows the distribution of loan officer per-formance, measured by the average percentage of correct decisions per session under the Baselinetreatment. The line plots the Kernel density of the performance distribution. We define a correctlending decision as approving an ex-post performing loan or declining an ex-post non-performing loan.

50

.4.5

.6.7

.8%

Len

ding

Dec

isio

ns C

orre

ct

0 5 10 15 20

Number of Experimental Sessions Completed

(a) Accuracy of Lending Decisions

−.2

0.2

.4.6

Pro

fit p

er A

ppro

ved

Loan

[U

S$

’000

]

0 5 10 15 20

Number of Experimental Sessions Completed

(b) Profitability of Lending Decisions

Figure 2: Learning During the Experiment. This figure examines the presence of learningeffects over the course of the experiment by plotting (a) the percentage of correct decisions by thetotal number of experimental sessions completed and (b) the profit per approved loan by the numberof experimental sessions completed. A correct lending decision is defined as a loan officer correctlyapproving a performing loan or correctly declining a loan that became delinquent. The dashed linesand shaded areas are Kernel-weighted local polynomial regressions and 95% confidence intervals.

51

0.0

1.0

2.0

3.0

4

20 30 40 50 60 70 80 90 100

(a) Non-performing Loans

0.0

1.0

2.0

3.0

4

20 30 40 50 60 70 80 90 100

(b) Performing Loans

Figure 3: Distribution of Internal Risk Ratings. This figure plots the distribution of internalratings assigned to loans evaluated under the baseline treatment. Panel (a) shows the distributionof risk-ratings for the sample of non-performing loans and loans that were declined by the Lenderex-ante; Panel (b) plots the distribution for performing loans. Vertical lines show the median of thedistribution. A Kolmogorov-Smirnov test rejects equality of distributions at 1% (p-value<0.001).

52